update app

.gitignore

@@ -0,0 +1,39 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+.ipynb_checkpoints/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+inference/
+inference_results/
+output/
+e2e_results/
+rec_results/
+det_results/
+train_data/
+log/
+*.DS_Store
+*.vs
+*.user
+*~
+*.vscode
+*.idea
+
+*.log
+.clang-format
+.clang_format.hook
+
+build/
+dist/
+paddleocr.egg-info/
+/deploy/android_demo/app/OpenCV/
+/deploy/android_demo/app/PaddleLite/
+/deploy/android_demo/app/.cxx/
+/deploy/android_demo/app/cache/
+test_tipc/web/models/
+test_tipc/web/node_modules/
+ppocr/
+.idea/

.pre-commit-config.yaml

@@ -0,0 +1,45 @@
+exclude: ^tools/utils/
+repos:
+  - repo: https://github.com/PyCQA/flake8
+    rev: 5.0.4
+    hooks:
+      - id: flake8
+        args: ["--max-line-length=150", "--ignore=E303,E731,W191,W504,E402,E722,E501,E203,E741,E265,E262,E126", "--exclude=__init__.py"]
+  - repo: https://github.com/pre-commit/mirrors-yapf
+    rev: v0.32.0
+    hooks:
+      - id: yapf
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.3.0
+    hooks:
+      - id: trailing-whitespace
+        exclude: |
+            (?x)^(
+              dicts/|
+              projects/.*?/dicts/
+            )
+      - id: check-yaml
+      - id: end-of-file-fixer
+        exclude: |
+            (?x)^(
+              dicts/|
+              projects/.*?/dicts/
+            )
+      - id: requirements-txt-fixer
+      - id: double-quote-string-fixer
+      - id: check-merge-conflict
+      - id: fix-encoding-pragma
+        args: ["--remove"]
+      - id: mixed-line-ending
+        args: ["--fix=lf"]
+      - id: mixed-line-ending
+        args: ["--fix=lf"]
+  - repo: https://github.com/executablebooks/mdformat
+    rev: 0.7.9
+    hooks:
+      - id: mdformat
+        args: ["--number", "--table-width", "200"]
+        additional_dependencies:
+          - mdformat-openmmlab
+          - mdformat_frontmatter
+          - linkify-it-py

LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

README.md

@@ -1,14 +1,283 @@
----
-title: OpenOCR UniRec Demo
-emoji: 🏢
-colorFrom: red
-colorTo: pink
-sdk: gradio
-sdk_version: 6.0.1
-app_file: app.py
-pinned: false
-license: apache-2.0
-short_description: Unified Text and Formula Recognition with 0.1B Parameters
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+<div align="center">
+
+<h1> OpenOCR: A general OCR system with accuracy and efficiency </h1>
+
+<h5 align="center"> If you find this project useful, please give us a star🌟. </h5>
+
+<a href="https://github.com/Topdu/OpenOCR/blob/main/LICENSE"><img alt="license" src="https://img.shields.io/github/license/Topdu/OpenOCR"></a>
+<a href='https://arxiv.org/abs/2411.15858'><img src='https://img.shields.io/badge/Paper-Arxiv-red'></a>
+<a href="https://huggingface.co/spaces/topdu/OpenOCR-Demo" target="_blank"><img src="https://img.shields.io/badge/%F0%9F%A4%97-Hugging Face Demo-blue"></a>
+<a href="https://modelscope.cn/studios/topdktu/OpenOCR-Demo" target="_blank"><img src="https://img.shields.io/badge/魔搭-Demo-blue"></a>
+<a href=""><img src="https://img.shields.io/badge/OS-Linux%2C%20Win%2C%20Mac-pink.svg"></a>
+<a href="https://github.com/Topdu/OpenOCR/graphs/contributors"><img src="https://img.shields.io/github/contributors/Topdu/OpenOCR?color=9ea"></a>
+<a href="https://pepy.tech/project/openocr"><img src="https://static.pepy.tech/personalized-badge/openocr?period=total&units=abbreviation&left_color=grey&right_color=blue&left_text=Clone%20downloads"></a>
+<a href="https://github.com/Topdu/OpenOCR/stargazers"><img src="https://img.shields.io/github/stars/Topdu/OpenOCR?color=ccf"></a>
+<a href="https://pypi.org/project/openocr-python/"><img alt="PyPI" src="https://img.shields.io/pypi/v/openocr-python"><img src="https://img.shields.io/pypi/dm/openocr-python?label=PyPI%20downloads"></a>
+
+<a href="#quick-start"> 🚀 Quick Start </a> | English | [简体中文](./README_ch.md)
+
+</div>
+
+______________________________________________________________________
+
+We aim to establish a unified benchmark for training and evaluating models in scene text detection and recognition. Building on this benchmark, we introduce a general OCR system with accuracy and efficiency, **OpenOCR**. This repository also serves as the official codebase of the OCR team from the [FVL Laboratory](https://fvl.fudan.edu.cn), Fudan University.
+
+We sincerely welcome the researcher to recommend OCR or relevant algorithms and point out any potential factual errors or bugs. Upon receiving the suggestions, we will promptly evaluate and critically reproduce them. We look forward to collaborating with you to advance the development of OpenOCR and continuously contribute to the OCR community!
+
+## Features
+
+- 🔥**UniRec: Unified Text and Formula Recognition Across Granularities**
+
+  - ⚡\[[Doc](./docs/unirec.md)\] \[[Model](https://huggingface.co/topdu/unirec_100m)\] \[[ModelScope Demo](https://www.modelscope.cn/studios/topdktu/OpenOCR-UniRec-Demo)\] \[[Hugging Face Demo](https://huggingface.co/spaces/topdu/OpenOCR-UniRec-Demo)\] \[[Local Demo](./docs/unirec.md#local-demo)\] \[Paper coming soon\]
+    - Recognizing plain text (words, lines, paragraphs), formulas (single-line, multi-line), and mixed text-and-formulas content.
+    - 0.1B parameters.
+    - Trained from scratch on 50M data without pre-training.
+
+- 🔥**OpenOCR: A general OCR system with accuracy and efficiency**
+
+  - ⚡\[[Quick Start](#quick-start)\] \[[Model](https://github.com/Topdu/OpenOCR/releases/tag/develop0.0.1)\] \[[ModelScope Demo](https://modelscope.cn/studios/topdktu/OpenOCR-Demo)\] \[[Hugging Face Demo](https://huggingface.co/spaces/topdu/OpenOCR-Demo)\] \[[Local Demo](#local-demo)\]  \[[PaddleOCR Implementation](https://paddlepaddle.github.io/PaddleOCR/latest/algorithm/text_recognition/algorithm_rec_svtrv2.html)\]
+  - [Introduction](./docs/openocr.md)
+    - A practical OCR system building on SVTRv2.
+    - Outperforms [PP-OCRv4](https://paddlepaddle.github.io/PaddleOCR/latest/ppocr/model_list.html) baseline by 4.5% on the [OCR competition leaderboard](https://aistudio.baidu.com/competition/detail/1131/0/leaderboard) in terms of accuracy, while preserving quite similar inference speed.
+    - [x] Supports Chinese and English text detection and recognition.
+    - [x] Provides server model and mobile model.
+    - [x] Fine-tunes OpenOCR on a custom dataset: [Fine-tuning Det](./docs/finetune_det.md), [Fine-tuning Rec](./docs/finetune_rec.md).
+    - [x] [ONNX model export for wider compatibility](#export-onnx-model).
+
+- 🔥**SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition (ICCV 2025)**
+
+  - \[[Paper](https://arxiv.org/abs/2411.15858)\] \[[Doc](./configs/rec/svtrv2/)\] \[[Model](./configs/rec/svtrv2/readme.md#11-models-and-results)\] \[[Datasets](./docs/svtrv2.md#downloading-datasets)\] \[[Config, Training and Inference](./configs/rec/svtrv2/readme.md#3-model-training--evaluation)\] \[[Benchmark](./docs/svtrv2.md#results-benchmark--configs--checkpoints)\]
+  - [Introduction](./docs/svtrv2.md)
+    - A unified training and evaluation benchmark (on top of [Union14M](https://github.com/Mountchicken/Union14M?tab=readme-ov-file#3-union14m-dataset)) for Scene Text Recognition
+    - Supports 24 Scene Text Recognition methods trained from scratch on the large-scale real dataset [Union14M-L-Filter](./docs/svtrv2.md#dataset-details), and will continue to add the latest methods.
+    - Improves accuracy by 20-30% compared to models trained based on synthetic datasets.
+    - Towards Arbitrary-Shaped Text Recognition and Language modeling with a Single Visual Model.
+    - Surpasses Attention-based Encoder-Decoder Methods across challenging scenarios in terms of accuracy and speed
+  - [Get Started](./docs/svtrv2.md#get-started-with-training-a-sota-scene-text-recognition-model-from-scratch) with training a SOTA Scene Text Recognition model from scratch.
+
+## Ours STR algorithms
+
+- [**SVTRv2**](./configs/rec/svtrv2) (*Yongkun Du, Zhineng Chen\*, Hongtao Xie, Caiyan Jia, Yu-Gang Jiang. SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition,* ICCV 2025. [Doc](./configs/rec/svtrv2/), [Paper](https://arxiv.org/abs/2411.15858))
+- [**IGTR**](./configs/rec/igtr/) (*Yongkun Du, Zhineng Chen\*, Yuchen Su, Caiyan Jia, Yu-Gang Jiang. Instruction-Guided Scene Text Recognition,* TPAMI 2025. [Doc](./configs/rec/igtr), [Paper](https://ieeexplore.ieee.org/document/10820836))
+- [**CPPD**](./configs/rec/cppd/) (*Yongkun Du, Zhineng Chen\*, Caiyan Jia, Xiaoting Yin, Chenxia Li, Yuning Du, Yu-Gang Jiang. Context Perception Parallel Decoder for Scene Text Recognition,* TPAMI 2025. [PaddleOCR Doc](https://github.com/PaddlePaddle/PaddleOCR/blob/main/docs/algorithm/text_recognition/algorithm_rec_cppd.en.md), [Paper](https://ieeexplore.ieee.org/document/10902187))
+- [**SMTR&FocalSVTR**](./configs/rec/smtr/) (*Yongkun Du, Zhineng Chen\*, Caiyan Jia, Xieping Gao, Yu-Gang Jiang. Out of Length Text Recognition with Sub-String Matching,* AAAI 2025. [Doc](./configs/rec/smtr/), [Paper](https://ojs.aaai.org/index.php/AAAI/article/view/32285))
+- [**DPTR**](./configs/rec/dptr/) (*Shuai Zhao, Yongkun Du, Zhineng Chen\*, Yu-Gang Jiang. Decoder Pre-Training with only Text for Scene Text Recognition,* ACM MM 2024. [Paper](https://dl.acm.org/doi/10.1145/3664647.3681390))
+- [**CDistNet**](./configs/rec/cdistnet/) (*Tianlun Zheng, Zhineng Chen\*, Shancheng Fang, Hongtao Xie, Yu-Gang Jiang. CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition,* IJCV 2024. [Paper](https://link.springer.com/article/10.1007/s11263-023-01880-0))
+- **MRN** (*Tianlun Zheng, Zhineng Chen\*, Bingchen Huang, Wei Zhang, Yu-Gang Jiang. MRN: Multiplexed Routing Network for Incremental Multilingual Text Recognition,* ICCV 2023. [Paper](https://openaccess.thecvf.com/content/ICCV2023/html/Zheng_MRN_Multiplexed_Routing_Network_for_Incremental_Multilingual_Text_Recognition_ICCV_2023_paper.html), [Code](https://github.com/simplify23/MRN))
+- **TPS++** (*Tianlun Zheng, Zhineng Chen\*, Jinfeng Bai, Hongtao Xie, Yu-Gang Jiang. TPS++: Attention-Enhanced Thin-Plate Spline for Scene Text Recognition,* IJCAI 2023. [Paper](https://arxiv.org/abs/2305.05322), [Code](https://github.com/simplify23/TPS_PP))
+- [**SVTR**](./configs/rec/svtr/) (*Yongkun Du, Zhineng Chen\*, Caiyan Jia, Xiaoting Yin, Tianlun Zheng, Chenxia Li, Yuning Du, Yu-Gang Jiang. SVTR: Scene Text Recognition with a Single Visual Model,* IJCAI 2022 (Long). [PaddleOCR Doc](https://github.com/Topdu/PaddleOCR/blob/main/doc/doc_ch/algorithm_rec_svtr.md), [Paper](https://www.ijcai.org/proceedings/2022/124))
+- [**NRTR**](./configs/rec/nrtr/) (*Fenfen Sheng, Zhineng Chen, Bo Xu. NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition,* ICDAR 2019. [Paper](https://arxiv.org/abs/1806.00926))
+
+## Recent Updates
+
+- **2025.07.10**: Our paper [SVTRv2](https://arxiv.org/abs/2411.15858) is accepted by ICCV 2025. Accessible in [Doc](./configs/rec/svtrv2/).
+
+- **2025.03.24**: 🔥 Releasing the feature of fine-tuning OpenOCR on a custom dataset: [Fine-tuning Det](./docs/finetune_det.md), [Fine-tuning Rec](./docs/finetune_rec.md)
+
+- **2025.03.23**: 🔥 Releasing the feature of [ONNX model export for wider compatibility](#export-onnx-model).
+
+- **2025.02.22**: Our paper [CPPD](https://ieeexplore.ieee.org/document/10902187) is accepted by TPAMI. Accessible in [Doc](./configs/rec/cppd/) and [PaddleOCR Doc](https://github.com/PaddlePaddle/PaddleOCR/blob/main/docs/algorithm/text_recognition/algorithm_rec_cppd.en.md).
+
+- **2024.12.31**: Our paper [IGTR](https://ieeexplore.ieee.org/document/10820836) is accepted by TPAMI. Accessible in [Doc](./configs/rec/igtr/).
+
+- **2024.12.16**: Our paper [SMTR](https://ojs.aaai.org/index.php/AAAI/article/view/32285) is accepted by AAAI 2025. Accessible in [Doc](./configs/rec/smtr/).
+
+- **2024.12.03**: The pre-training code for [DPTR](https://dl.acm.org/doi/10.1145/3664647.3681390) is merged.
+
+- **🔥 2024.11.23 release notes**:
+
+  - **OpenOCR: A general OCR system with accuracy and efficiency**
+    - ⚡\[[Quick Start](#quick-start)\] \[[Model](https://github.com/Topdu/OpenOCR/releases/tag/develop0.0.1)\] \[[ModelScope Demo](https://modelscope.cn/studios/topdktu/OpenOCR-Demo)\] \[[Hugging Face Demo](https://huggingface.co/spaces/topdu/OpenOCR-Demo)\] \[[Local Demo](#local-demo)\]  \[[PaddleOCR Implementation](https://paddlepaddle.github.io/PaddleOCR/latest/algorithm/text_recognition/algorithm_rec_svtrv2.html)\]
+    - [Introduction](./docs/openocr.md)
+  - **SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition**
+    - \[[Paper](https://arxiv.org/abs/2411.15858)\] \[[Doc](./configs/rec/svtrv2/)\] \[[Model](./configs/rec/svtrv2/readme.md#11-models-and-results)\] \[[Datasets](./docs/svtrv2.md#downloading-datasets)\] \[[Config, Training and Inference](./configs/rec/svtrv2/readme.md#3-model-training--evaluation)\] \[[Benchmark](./docs/svtrv2.md#results--configs--checkpoints)\]
+    - [Introduction](./docs/svtrv2.md)
+    - [Get Started](./docs/svtrv2.md#get-started-with-training-a-sota-scene-text-recognition-model-from-scratch) with training a SOTA Scene Text Recognition model from scratch.
+
+## Quick Start
+
+**Note**: OpenOCR supports inference using both the ONNX and Torch frameworks, with the dependency environments for the two frameworks being isolated. When using ONNX for inference, there is no need to install Torch, and vice versa.
+
+### 1. ONNX Inference
+
+#### Install OpenOCR and Dependencies:
+
+```shell
+pip install openocr-python
+pip install onnxruntime
+```
+
+#### Usage:
+
+```python
+from openocr import OpenOCR
+onnx_engine = OpenOCR(backend='onnx', device='cpu')
+img_path = '/path/img_path or /path/img_file'
+result, elapse = onnx_engine(img_path)
+```
+
+### 2. Pytorch inference
+
+#### Dependencies:
+
+- [PyTorch](http://pytorch.org/) version >= 1.13.0
+- Python version >= 3.7
+
+```shell
+conda create -n openocr python==3.8
+conda activate openocr
+# install gpu version torch
+conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=11.8 -c pytorch -c nvidia
+# or cpu version
+conda install pytorch torchvision torchaudio cpuonly -c pytorch
+```
+
+After installing dependencies, the following two installation methods are available. Either one can be chosen.
+
+#### 2.1. Python Modules
+
+**Install OpenOCR**:
+
+```shell
+pip install openocr-python
+```
+
+**Usage**:
+
+```python
+from openocr import OpenOCR
+engine = OpenOCR()
+img_path = '/path/img_path or /path/img_file'
+result, elapse = engine(img_path)
+
+# Server mode
+# engine = OpenOCR(mode='server')
+```
+
+#### 2.2. Clone this repository:
+
+```shell
+git clone https://github.com/Topdu/OpenOCR.git
+cd OpenOCR
+pip install -r requirements.txt
+wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/openocr_det_repvit_ch.pth
+wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/openocr_repsvtr_ch.pth
+# Rec Server model
+# wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/openocr_svtrv2_ch.pth
+```
+
+**Usage**:
+
+```shell
+# OpenOCR system: Det + Rec model
+python tools/infer_e2e.py --img_path=/path/img_fold or /path/img_file
+# Det model
+python tools/infer_det.py --c ./configs/det/dbnet/repvit_db.yml --o Global.infer_img=/path/img_fold or /path/img_file
+# Rec model
+python tools/infer_rec.py --c ./configs/rec/svtrv2/repsvtr_ch.yml --o Global.infer_img=/path/img_fold or /path/img_file
+```
+
+##### Export ONNX model
+
+```shell
+pip install onnx
+python tools/toonnx.py --c configs/rec/svtrv2/repsvtr_ch.yml --o Global.device=cpu
+python tools/toonnx.py --c configs/det/dbnet/repvit_db.yml --o Global.device=cpu
+```
+
+##### Inference with ONNXRuntime
+
+```shell
+pip install onnxruntime
+# OpenOCR system: Det + Rec model
+python tools/infer_e2e.py --img_path=/path/img_fold or /path/img_file --backend=onnx --device=cpu
+# Det model
+python tools/infer_det.py --c ./configs/det/dbnet/repvit_db.yml --o Global.backend=onnx Global.device=cpu Global.infer_img=/path/img_fold or /path/img_file
+# Rec model
+python tools/infer_rec.py --c ./configs/rec/svtrv2/repsvtr_ch.yml --o Global.backend=onnx Global.device=cpu Global.infer_img=/path/img_fold or /path/img_file
+```
+
+#### Local Demo
+
+```shell
+pip install gradio==4.20.0
+wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/OCR_e2e_img.tar
+tar xf OCR_e2e_img.tar
+# start demo
+python demo_gradio.py
+```
+
+## Reproduction schedule:
+
+### Scene Text Recognition
+
+| Method                                        | Venue                                                                                          | Training | Evaluation | Contributor                                 |
+| --------------------------------------------- | ---------------------------------------------------------------------------------------------- | -------- | ---------- | ------------------------------------------- |
+| [CRNN](./configs/rec/svtrs/)                  | [TPAMI 2016](https://arxiv.org/abs/1507.05717)                                                 | ✅       | ✅         |                                             |
+| [ASTER](./configs/rec/aster/)                 | [TPAMI 2019](https://ieeexplore.ieee.org/document/8395027)                                     | ✅       | ✅         | [pretto0](https://github.com/pretto0)       |
+| [NRTR](./configs/rec/nrtr/)                   | [ICDAR 2019](https://arxiv.org/abs/1806.00926)                                                 | ✅       | ✅         |                                             |
+| [SAR](./configs/rec/sar/)                     | [AAAI 2019](https://aaai.org/papers/08610-show-attend-and-read-a-simple-and-strong-baseline-for-irregular-text-recognition/) | ✅       | ✅         | [pretto0](https://github.com/pretto0)       |
+| [MORAN](./configs/rec/moran/)                 | [PR 2019](https://www.sciencedirect.com/science/article/abs/pii/S0031320319300263)             | ✅       | ✅         |                                             |
+| [DAN](./configs/rec/dan/)                     | [AAAI 2020](https://arxiv.org/pdf/1912.10205)                                                  | ✅       | ✅         |                                             |
+| [RobustScanner](./configs/rec/robustscanner/) | [ECCV 2020](https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/3160_ECCV_2020_paper.php)   | ✅       | ✅         | [pretto0](https://github.com/pretto0)       |
+| [AutoSTR](./configs/rec/autostr/)             | [ECCV 2020](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123690732.pdf)            | ✅       | ✅         |                                             |
+| [SRN](./configs/rec/srn/)                     | [CVPR 2020](https://openaccess.thecvf.com/content_CVPR_2020/html/Yu_Towards_Accurate_Scene_Text_Recognition_With_Semantic_Reasoning_Networks_CVPR_2020_paper.html) | ✅       | ✅         | [pretto0](https://github.com/pretto0)       |
+| [SEED](./configs/rec/seed/)                   | [CVPR 2020](https://openaccess.thecvf.com/content_CVPR_2020/html/Qiao_SEED_Semantics_Enhanced_Encoder-Decoder_Framework_for_Scene_Text_Recognition_CVPR_2020_paper.html) | ✅       | ✅         |                                             |
+| [ABINet](./configs/rec/abinet/)               | [CVPR 2021](https://openaccess.thecvf.com//content/CVPR2021/html/Fang_Read_Like_Humans_Autonomous_Bidirectional_and_Iterative_Language_Modeling_for_CVPR_2021_paper.html) | ✅       | ✅         | [YesianRohn](https://github.com/YesianRohn) |
+| [VisionLAN](./configs/rec/visionlan/)         | [ICCV 2021](https://openaccess.thecvf.com/content/ICCV2021/html/Wang_From_Two_to_One_A_New_Scene_Text_Recognizer_With_ICCV_2021_paper.html) | ✅       | ✅         | [YesianRohn](https://github.com/YesianRohn) |
+| PIMNet                                        | [ACM MM 2021](https://dl.acm.org/doi/10.1145/3474085.3475238)                                  |          |            | TODO                                        |
+| [SVTR](./configs/rec/svtrs/)                  | [IJCAI 2022](https://www.ijcai.org/proceedings/2022/124)                                       | ✅       | ✅         |                                             |
+| [PARSeq](./configs/rec/parseq/)               | [ECCV 2022](https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880177.pdf)            | ✅       | ✅         |                                             |
+| [MATRN](./configs/rec/matrn/)                 | [ECCV 2022](https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880442.pdf)            | ✅       | ✅         |                                             |
+| [MGP-STR](./configs/rec/mgpstr/)              | [ECCV 2022](https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880336.pdf)            | ✅       | ✅         |                                             |
+| [LPV](./configs/rec/lpv/)                     | [IJCAI 2023](https://www.ijcai.org/proceedings/2023/0189.pdf)                                  | ✅       | ✅         |                                             |
+| [MAERec](./configs/rec/maerec/)(Union14M)     | [ICCV 2023](https://openaccess.thecvf.com/content/ICCV2023/papers/Jiang_Revisiting_Scene_Text_Recognition_A_Data_Perspective_ICCV_2023_paper.pdf) | ✅       | ✅         |                                             |
+| [LISTER](./configs/rec/lister/)               | [ICCV 2023](https://openaccess.thecvf.com/content/ICCV2023/papers/Cheng_LISTER_Neighbor_Decoding_for_Length-Insensitive_Scene_Text_Recognition_ICCV_2023_paper.pdf) | ✅       | ✅         |                                             |
+| [CDistNet](./configs/rec/cdistnet/)           | [IJCV 2024](https://link.springer.com/article/10.1007/s11263-023-01880-0)                      | ✅       | ✅         | [YesianRohn](https://github.com/YesianRohn) |
+| [BUSNet](./configs/rec/busnet/)               | [AAAI 2024](https://ojs.aaai.org/index.php/AAAI/article/view/28402)                            | ✅       | ✅         |                                             |
+| DCTC                                          | [AAAI 2024](https://ojs.aaai.org/index.php/AAAI/article/view/28575)                            |          |            | TODO                                        |
+| [CAM](./configs/rec/cam/)                     | [PR 2024](https://arxiv.org/abs/2402.13643)                                                    | ✅       | ✅         |                                             |
+| [OTE](./configs/rec/ote/)                     | [CVPR 2024](https://openaccess.thecvf.com/content/CVPR2024/html/Xu_OTE_Exploring_Accurate_Scene_Text_Recognition_Using_One_Token_CVPR_2024_paper.html) | ✅       | ✅         |                                             |
+| CFF                                           | [IJCAI 2024](https://arxiv.org/abs/2407.05562)                                                 |          |            | TODO                                        |
+| [DPTR](./configs/rec/dptr/)                   | [ACM MM 2024](https://dl.acm.org/doi/10.1145/3664647.3681390)                                  |          |            | [fd-zs](https://github.com/fd-zs)           |
+| VIPTR                                         | [ACM CIKM 2024](https://arxiv.org/abs/2401.10110)                                              |          |            | TODO                                        |
+| [IGTR](./configs/rec/igtr/)                   | [TPAMI 2025](https://ieeexplore.ieee.org/document/10820836)                                    | ✅       | ✅         |                                             |
+| [SMTR](./configs/rec/smtr/)                   | [AAAI 2025](https://ojs.aaai.org/index.php/AAAI/article/view/32285)                            | ✅       | ✅         |                                             |
+| [CPPD](./configs/rec/cppd/)                   | [TPAMI 2025](https://ieeexplore.ieee.org/document/10902187)                                    | ✅       | ✅         |                                             |
+| [FocalSVTR-CTC](./configs/rec/svtrs/)         | [AAAI 2025](https://ojs.aaai.org/index.php/AAAI/article/view/32285)                            | ✅       | ✅         |                                             |
+| [SVTRv2](./configs/rec/svtrv2/)               | [ICCV 2025](https://arxiv.org/abs/2411.15858)                                                  | ✅       | ✅         |                                             |
+| [ResNet+Trans-CTC](./configs/rec/svtrs/)      |                                                                                                | ✅       | ✅         |                                             |
+| [ViT-CTC](./configs/rec/svtrs/)               |                                                                                                | ✅       | ✅         |                                             |
+
+#### Contributors
+
+______________________________________________________________________
+
+Yiming Lei ([pretto0](https://github.com/pretto0)), Xingsong Ye ([YesianRohn](https://github.com/YesianRohn)), and Shuai Zhao ([fd-zs](https://github.com/fd-zs)) from the [FVL Laboratory](https://fvl.fudan.edu.cn), Fudan University, with guidance from Dr. Zhineng Chen ([Homepage](https://zhinchenfd.github.io/)), completed the majority work of the algorithm reproduction. Grateful for their outstanding contributions.
+
+### Scene Text Detection (STD)
+
+TODO
+
+### Text Spotting
+
+TODO
+
+______________________________________________________________________
+
+## Citation
+
+If you find our method useful for your reserach, please cite:
+
+```bibtex
+@inproceedings{Du2024SVTRv2,
+      title={SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition},
+      author={Yongkun Du and Zhineng Chen and Hongtao Xie and Caiyan Jia and Yu-Gang Jiang},
+      booktitle={ICCV},
+      year={2025}
+}
+```
+
+# Acknowledgement
+
+This codebase is built based on the [PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR), [PytorchOCR](https://github.com/WenmuZhou/PytorchOCR), and [MMOCR](https://github.com/open-mmlab/mmocr). Thanks for their awesome work!

README_ch.md

@@ -0,0 +1,273 @@
+<div align="center">
+
+<h1> OpenOCR: A general OCR system with accuracy and efficiency </h1>
+
+<h5 align="center"> 如果您觉得本项目有帮助，请为我们点亮Star🌟 </h5>
+
+<a href="https://github.com/Topdu/OpenOCR/blob/main/LICENSE"><img alt="license" src="https://img.shields.io/github/license/Topdu/OpenOCR"></a>
+<a href='https://arxiv.org/abs/2411.15858'><img src='https://img.shields.io/badge/论文-Arxiv-red'></a>
+<a href="https://huggingface.co/spaces/topdu/OpenOCR-Demo" target="_blank"><img src="https://img.shields.io/badge/%F0%9F%A4%97-Hugging Face Demo-blue"></a>
+<a href="https://modelscope.cn/studios/topdktu/OpenOCR-Demo" target="_blank"><img src="https://img.shields.io/badge/魔搭-Demo-blue"></a>
+<a href=""><img src="https://img.shields.io/badge/OS-Linux%2C%20Win%2C%20Mac-pink.svg"></a>
+<a href="https://github.com/Topdu/OpenOCR/graphs/contributors"><img src="https://img.shields.io/github/contributors/Topdu/OpenOCR?color=9ea"></a>
+<a href="https://pepy.tech/project/openocr"><img src="https://static.pepy.tech/personalized-badge/openocr?period=total&units=abbreviation&left_color=grey&right_color=blue&left_text=Clone%20下载量"></a>
+<a href="https://github.com/Topdu/OpenOCR/stargazers"><img src="https://img.shields.io/github/stars/Topdu/OpenOCR?color=ccf"></a>
+<a href="https://pypi.org/project/openocr-python/"><img alt="PyPI" src="https://img.shields.io/pypi/v/openocr-python"><img src="https://img.shields.io/pypi/dm/openocr-python?label=PyPI%20下载量"></a>
+
+<a href="#快速开始"> 🚀 快速开始 </a> | 简体中文 | [English](./README.md)
+
+</div>
+
+______________________________________________________________________
+
+我们致力于构建场景文本检测与识别模型的统一训练评估基准。基于此基准，我们推出了兼顾精度与效率的通用OCR系统——**OpenOCR**。本仓库同时作为复旦大学[FVL实验室](https://fvl.fudan.edu.cn)OCR团队的官方代码库。
+
+我们诚挚欢迎研究者推荐OCR相关算法，并指出潜在的事实性错误或代码缺陷。收到建议后，我们将及时评估并严谨复现。期待与您携手推进OpenOCR发展，持续为OCR社区贡献力量！
+
+## 核心特性
+
+- 🔥**UniRec: Unified Text and Formula Recognition Across Granularities**
+
+  - ⚡\[[使用文档](./docs/unirec.md)\] \[[模型下载](https://huggingface.co/topdu/unirec_100m)\] \[[ModelScope Demo](https://www.modelscope.cn/studios/topdktu/OpenOCR-UniRec-Demo)\] \[[Hugging Face Demo](https://huggingface.co/spaces/topdu/OpenOCR-UniRec-Demo)\] \[[本地Demo](./docs/unirec.md#local-demo)\] \[论文coming soon\]
+    - 识别纯文本（单词、行、段落）、公式（单行、多行）、以及文本与公式混合的内容
+    - 0.1B 参数量
+    - 在 5000 万数据上从零开始训练，不使用任何预训练
+
+- 🔥**OpenOCR: A general OCR system with accuracy and efficiency**
+
+  - ⚡\[[快速开始](#快速开始)\] \[[模型下载](https://github.com/Topdu/OpenOCR/releases/tag/develop0.0.1)\] \[[ModelScope Demo](https://modelscope.cn/studios/topdktu/OpenOCR-Demo)\] \[[Hugging Face Demo](https://huggingface.co/spaces/topdu/OpenOCR-Demo)\] \[[本地Demo](#本地Demo)\] \[[PaddleOCR实现](https://paddlepaddle.github.io/PaddleOCR/latest/algorithm/text_recognition/algorithm_rec_svtrv2.html)\]
+  - [技术文档](./docs/openocr.md)
+    - 基于SVTRv2构建的实用OCR系统
+    - 在[OCR竞赛榜单](https://aistudio.baidu.com/competition/detail/1131/0/leaderboard)上，精度超越[PP-OCRv4](https://paddlepaddle.github.io/PaddleOCR/latest/ppocr/model_list.html)基线4.5%，推理速度保持相近
+    - [x] 支持中英文文本检测与识别
+    - [x] 提供服务器端(Server)与移动端(mobile)模型
+    - [x] 支持自定义数据集微调: [检测模型微调](./docs/finetune_det.md), [识别模型微调](./docs/finetune_rec.md)
+    - [x] [支持导出ONNX模型](#导出onnx模型)
+
+- 🔥**SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition (ICCV 2025)**
+
+  - \[[论文](https://arxiv.org/abs/2411.15858)\] \[[文档](./configs/rec/svtrv2/)\] \[[模型](./configs/rec/svtrv2/readme.md#11-models-and-results)\] \[[数据集](./docs/svtrv2.md#downloading-datasets)\] \[[配置/训练/推理](./configs/rec/svtrv2/readme.md#3-model-training--evaluation)\] \[[基准测试](./docs/svtrv2.md#results-benchmark--configs--checkpoints)\]
+  - [技术文档](./docs/svtrv2.md)
+    - 基于[Union14M](https://github.com/Mountchicken/Union14M)构建的场景文本识别统一训练评估基准
+    - 支持24种场景文本识别方法在大规模真实数据集[Union14M-L-Filter](./docs/svtrv2.md#数据集详情)上的训练，将持续集成前沿方法
+    - 相比基于合成数据训练的模型，精度提升20-30%
+    - 单一视觉模型实现任意形状文本识别与语言建模
+    - 在精度与速度上全面超越基于Attention的编解码模型
+    - [从零训练SOTA模型指南](./docs/svtrv2.md#get-started-with-training-a-sota-scene-text-recognition-model-from-scratch)
+
+## 自研STR算法
+
+- [**SVTRv2**](./configs/rec/svtrv2) (*Yongkun Du, Zhineng Chen\*, Hongtao Xie, Caiyan Jia, Yu-Gang Jiang. SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition,* ICCV 2025. [Doc](./configs/rec/svtrv2/), [Paper](https://arxiv.org/abs/2411.15858))
+- [**IGTR**](./configs/rec/igtr/) (*Yongkun Du, Zhineng Chen\*, Yuchen Su, Caiyan Jia, Yu-Gang Jiang. Instruction-Guided Scene Text Recognition,* TPAMI 2025. [Doc](./configs/rec/igtr), [Paper](https://ieeexplore.ieee.org/document/10820836))
+- [**CPPD**](./configs/rec/cppd/) (*Yongkun Du, Zhineng Chen\*, Caiyan Jia, Xiaoting Yin, Chenxia Li, Yuning Du, Yu-Gang Jiang. Context Perception Parallel Decoder for Scene Text Recognition,* TPAMI 2025. [PaddleOCR Doc](https://github.com/PaddlePaddle/PaddleOCR/blob/main/docs/algorithm/text_recognition/algorithm_rec_cppd.en.md), [Paper](https://ieeexplore.ieee.org/document/10902187))
+- [**SMTR&FocalSVTR**](./configs/rec/smtr/) (*Yongkun Du, Zhineng Chen\*, Caiyan Jia, Xieping Gao, Yu-Gang Jiang. Out of Length Text Recognition with Sub-String Matching,* AAAI 2025. [Doc](./configs/rec/smtr/), [Paper](https://ojs.aaai.org/index.php/AAAI/article/view/32285))
+- [**DPTR**](./configs/rec/dptr/) (*Shuai Zhao, Yongkun Du, Zhineng Chen\*, Yu-Gang Jiang. Decoder Pre-Training with only Text for Scene Text Recognition,* ACM MM 2024. [Paper](https://dl.acm.org/doi/10.1145/3664647.3681390))
+- [**CDistNet**](./configs/rec/cdistnet/) (*Tianlun Zheng, Zhineng Chen\*, Shancheng Fang, Hongtao Xie, Yu-Gang Jiang. CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition,* IJCV 2024. [Paper](https://link.springer.com/article/10.1007/s11263-023-01880-0))
+- **MRN** (*Tianlun Zheng, Zhineng Chen\*, Bingchen Huang, Wei Zhang, Yu-Gang Jiang. MRN: Multiplexed Routing Network for Incremental Multilingual Text Recognition,* ICCV 2023. [Paper](https://openaccess.thecvf.com/content/ICCV2023/html/Zheng_MRN_Multiplexed_Routing_Network_for_Incremental_Multilingual_Text_Recognition_ICCV_2023_paper.html), [Code](https://github.com/simplify23/MRN))
+- **TPS++** (*Tianlun Zheng, Zhineng Chen\*, Jinfeng Bai, Hongtao Xie, Yu-Gang Jiang. TPS++: Attention-Enhanced Thin-Plate Spline for Scene Text Recognition,* IJCAI 2023. [Paper](https://arxiv.org/abs/2305.05322), [Code](https://github.com/simplify23/TPS_PP))
+- [**SVTR**](./configs/rec/svtr/) (*Yongkun Du, Zhineng Chen\*, Caiyan Jia, Xiaoting Yin, Tianlun Zheng, Chenxia Li, Yuning Du, Yu-Gang Jiang. SVTR: Scene Text Recognition with a Single Visual Model,* IJCAI 2022 (Long). [PaddleOCR Doc](https://github.com/Topdu/PaddleOCR/blob/main/doc/doc_ch/algorithm_rec_svtr.md), [Paper](https://www.ijcai.org/proceedings/2022/124))
+- [**NRTR**](./configs/rec/nrtr/) (*Fenfen Sheng, Zhineng Chen, Bo Xu. NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition,* ICDAR 2019. [Paper](https://arxiv.org/abs/1806.00926))
+
+## 近期更新
+
+- **2025.07.10**: [SVTRv2](https://arxiv.org/abs/2411.15858)被ICCV 2025接收. 详见[文档](./configs/rec/svtrv2/)
+- **2025.03.24**: 🔥 发布自定义数据集微调功能: [检测模型微调](./docs/finetune_det.md), [识别模型微调](./docs/finetune_rec.md)
+- **2025.03.23**: 🔥 新增[ONNX模型导出功能](#导出onnx模型)
+- **2025.02.22**: [CPPD](https://ieeexplore.ieee.org/document/10902187)论文被TPAMI录用，详见[文档](./configs/rec/cppd/)与[PaddleOCR文档](https://github.com/PaddlePaddle/PaddleOCR/blob/main/docs/algorithm/text_recognition/algorithm_rec_cppd.en.md)
+- **2024.12.31**: [IGTR](https://ieeexplore.ieee.org/document/10820836)论文被TPAMI录用，详见[文档](./configs/rec/igtr/)
+- **2024.12.16**: [SMTR](https://ojs.aaai.org/index.php/AAAI/article/view/32285)论文被AAAI 2025录用，详见[文档](./configs/rec/smtr/)
+- **2024.12.03**: [DPTR](https://dl.acm.org/doi/10.1145/3664647.3681390)预训练代码合并
+- **🔥 2024.11.23 重大更新**:
+  - **OpenOCR通用OCR系统发布**
+    - ⚡\[[快速开始](#快速开始)\] \[[模型下载](https://github.com/Topdu/OpenOCR/releases/tag/develop0.0.1)\] \[[ModelScopeDemo](https://modelscope.cn/studios/topdktu/OpenOCR-Demo)\] \[[Hugging FaceDemo](https://huggingface.co/spaces/topdu/OpenOCR-Demo)\] \[[本地Demo](#本地Demo)\] \[[PaddleOCR实现](https://paddlepaddle.github.io/PaddleOCR/latest/algorithm/text_recognition/algorithm_rec_svtrv2.html)\]
+    - [技术文档](./docs/openocr.md)
+  - **SVTRv2论文发布**
+    - \[[论文](https://arxiv.org/abs/2411.15858)\] \[[文档](./configs/rec/svtrv2/)\] \[[模型](./configs/rec/svtrv2/readme.md#11-models-and-results)\] \[[数据集](./docs/svtrv2.md#downloading-datasets)\] \[[配置/训练/推理](./configs/rec/svtrv2/readme.md#3-model-training--evaluation)\] \[[基准测试](./docs/svtrv2.md#results-benchmark--configs--checkpoints)\]
+    - [技术文档](./docs/svtrv2.md)
+    - [从零训练SOTA模型指南](./docs/svtrv2.md#get-started-with-training-a-sota-scene-text-recognition-model-from-scratch)
+
+## 快速开始
+
+**注意**: OpenOCR支持ONNX和PyTorch双框架推理，环境相互独立。使用ONNX推理时无需安装PyTorch，反之亦然。
+
+### 1. ONNX推理
+
+#### 安装OpenOCR及依赖:
+
+```shell
+pip install openocr-python
+pip install onnxruntime
+```
+
+#### 使用示例:
+
+```python
+from openocr import OpenOCR
+onnx_engine = OpenOCR(backend='onnx', device='cpu')
+img_path = '/path/img_path or /path/img_file'
+result, elapse = onnx_engine(img_path)
+```
+
+### 2. PyTorch推理
+
+#### 环境依赖:
+
+- [PyTorch](http://pytorch.org/) >= 1.13.0
+- Python >= 3.7
+
+```shell
+conda create -n openocr python==3.8
+conda activate openocr
+# 安装GPU版本
+conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=11.8 -c pytorch -c nvidia
+# 或CPU版本
+conda install pytorch torchvision torchaudio cpuonly -c pytorch
+```
+
+#### 2.1 Python包安装
+
+**安装OpenOCR**:
+
+```shell
+pip install openocr-python
+```
+
+**使用示例**:
+
+```python
+from openocr import OpenOCR
+engine = OpenOCR()
+img_path = '/path/img_path or /path/img_file'
+result, elapse = engine(img_path)
+
+# Server模式
+# engine = OpenOCR(mode='server')
+```
+
+#### 2.2 源码安装
+
+```shell
+git clone https://github.com/Topdu/OpenOCR.git
+cd OpenOCR
+pip install -r requirements.txt
+wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/openocr_det_repvit_ch.pth
+wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/openocr_repsvtr_ch.pth
+# Server识别模型
+# wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/openocr_svtrv2_ch.pth
+```
+
+**使用命令**:
+
+```shell
+# 端到端OCR系统: 检测+识别
+python tools/infer_e2e.py --img_path=/path/img_path or /path/img_file
+# 单独检测模型
+python tools/infer_det.py --c ./configs/det/dbnet/repvit_db.yml --o Global.infer_img=/path/img_path or /path/img_file
+# 单独识别模型
+python tools/infer_rec.py --c ./configs/rec/svtrv2/repsvtr_ch.yml --o Global.infer_img=/path/img_path or /path/img_file
+```
+
+##### 导出ONNX模型
+
+```shell
+pip install onnx
+python tools/toonnx.py --c configs/rec/svtrv2/repsvtr_ch.yml --o Global.device=cpu
+python tools/toonnx.py --c configs/det/dbnet/repvit_db.yml --o Global.device=cpu
+```
+
+##### ONNXRuntime推理
+
+```shell
+pip install onnxruntime
+# 端到端OCR系统
+python tools/infer_e2e.py --img_path=/path/img_path or /path/img_file --backend=onnx --device=cpu
+# 检测模型
+python tools/infer_det.py --c ./configs/det/dbnet/repvit_db.yml --o Global.backend=onnx Global.device=cpu Global.infer_img=/path/img_path or /path/img_file
+# 识别模型
+python tools/infer_rec.py --c ./configs/rec/svtrv2/repsvtr_ch.yml --o Global.backend=onnx Global.device=cpu Global.infer_img=/path/img_path or /path/img_file
+```
+
+#### 本地Demo
+
+```shell
+pip install gradio==4.20.0
+wget https://github.com/Topdu/OpenOCR/releases/download/develop0.0.1/OCR_e2e_img.tar
+tar xf OCR_e2e_img.tar
+# 启动Demo
+python demo_gradio.py
+```
+
+## 算法复现计划
+
+### 场景文本识别(STR)
+
+| 方法                                          | 会议/期刊                                                                                        | 训练支持 | 评估支持 | 贡献者                                      |
+| --------------------------------------------- | ------------------------------------------------------------------------------------------------ | -------- | -------- | ------------------------------------------- |
+| [CRNN](./configs/rec/svtrs/)                  | [TPAMI 2016](https://arxiv.org/abs/1507.05717)                                                   | ✅       | ✅       |                                             |
+| [ASTER](./configs/rec/aster/)                 | [TPAMI 2019](https://ieeexplore.ieee.org/document/8395027)                                       | ✅       | ✅       | [pretto0](https://github.com/pretto0)       |
+| [NRTR](./configs/rec/nrtr/)                   | [ICDAR 2019](https://arxiv.org/abs/1806.00926)                                                   | ✅       | ✅       |                                             |
+| [SAR](./configs/rec/sar/)                     | [AAAI 2019](https://aaai.org/papers/08610-show-attend-and-read-a-simple-and-strong-baseline-for-irregular-text-recognition/) | ✅       | ✅       | [pretto0](https://github.com/pretto0)       |
+| [MORAN](./configs/rec/moran/)                 | [PR 2019](https://www.sciencedirect.com/science/article/abs/pii/S0031320319300263)               | ✅       | ✅       |                                             |
+| [DAN](./configs/rec/dan/)                     | [AAAI 2020](https://arxiv.org/pdf/1912.10205)                                                    | ✅       | ✅       |                                             |
+| [RobustScanner](./configs/rec/robustscanner/) | [ECCV 2020](https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/3160_ECCV_2020_paper.php)     | ✅       | ✅       | [pretto0](https://github.com/pretto0)       |
+| [AutoSTR](./configs/rec/autostr/)             | [ECCV 2020](https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123690732.pdf)              | ✅       | ✅       |                                             |
+| [SRN](./configs/rec/srn/)                     | [CVPR 2020](https://openaccess.thecvf.com/content_CVPR_2020/html/Yu_Towards_Accurate_Scene_Text_Recognition_With_Semantic_Reasoning_Networks_CVPR_2020_paper.html) | ✅       | ✅       | [pretto0](https://github.com/pretto0)       |
+| [SEED](./configs/rec/seed/)                   | [CVPR 2020](https://openaccess.thecvf.com/content_CVPR_2020/html/Qiao_SEED_Semantics_Enhanced_Encoder-Decoder_Framework_for_Scene_Text_Recognition_CVPR_2020_paper.html) | ✅       | ✅       |                                             |
+| [ABINet](./configs/rec/abinet/)               | [CVPR 2021](https://openaccess.thecvf.com//content/CVPR2021/html/Fang_Read_Like_Humans_Autonomous_Bidirectional_and_Iterative_Language_Modeling_for_CVPR_2021_paper.html) | ✅       | ✅       | [YesianRohn](https://github.com/YesianRohn) |
+| [VisionLAN](./configs/rec/visionlan/)         | [ICCV 2021](https://openaccess.thecvf.com/content/ICCV2021/html/Wang_From_Two_to_One_A_New_Scene_Text_Recognizer_With_ICCV_2021_paper.html) | ✅       | ✅       | [YesianRohn](https://github.com/YesianRohn) |
+| PIMNet                                        | [ACM MM 2021](https://dl.acm.org/doi/10.1145/3474085.3475238)                                    |          |          | TODO                                        |
+| [SVTR](./configs/rec/svtrs/)                  | [IJCAI 2022](https://www.ijcai.org/proceedings/2022/124)                                         | ✅       | ✅       |                                             |
+| [PARSeq](./configs/rec/parseq/)               | [ECCV 2022](https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880177.pdf)              | ✅       | ✅       |                                             |
+| [MATRN](./configs/rec/matrn/)                 | [ECCV 2022](https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880442.pdf)              | ✅       | ✅       |                                             |
+| [MGP-STR](./configs/rec/mgpstr/)              | [ECCV 2022](https://www.ecva.net/papers/eccv_2022/papers_ECCV/papers/136880336.pdf)              | ✅       | ✅       |                                             |
+| [LPV](./configs/rec/lpv/)                     | [IJCAI 2023](https://www.ijcai.org/proceedings/2023/0189.pdf)                                    | ✅       | ✅       |                                             |
+| [MAERec](./configs/rec/maerec/)(Union14M)     | [ICCV 2023](https://openaccess.thecvf.com/content/ICCV2023/papers/Jiang_Revisiting_Scene_Text_Recognition_A_Data_Perspective_ICCV_2023_paper.pdf) | ✅       | ✅       |                                             |
+| [LISTER](./configs/rec/lister/)               | [ICCV 2023](https://openaccess.thecvf.com/content/ICCV2023/papers/Cheng_LISTER_Neighbor_Decoding_for_Length-Insensitive_Scene_Text_Recognition_ICCV_2023_paper.pdf) | ✅       | ✅       |                                             |
+| [CDistNet](./configs/rec/cdistnet/)           | [IJCV 2024](https://link.springer.com/article/10.1007/s11263-023-01880-0)                        | ✅       | ✅       | [YesianRohn](https://github.com/YesianRohn) |
+| [BUSNet](./configs/rec/busnet/)               | [AAAI 2024](https://ojs.aaai.org/index.php/AAAI/article/view/28402)                              | ✅       | ✅       |                                             |
+| DCTC                                          | [AAAI 2024](https://ojs.aaai.org/index.php/AAAI/article/view/28575)                              |          |          | TODO                                        |
+| [CAM](./configs/rec/cam/)                     | [PR 2024](https://arxiv.org/abs/2402.13643)                                                      | ✅       | ✅       |                                             |
+| [OTE](./configs/rec/ote/)                     | [CVPR 2024](https://openaccess.thecvf.com/content/CVPR2024/html/Xu_OTE_Exploring_Accurate_Scene_Text_Recognition_Using_One_Token_CVPR_2024_paper.html) | ✅       | ✅       |                                             |
+| CFF                                           | [IJCAI 2024](https://arxiv.org/abs/2407.05562)                                                   |          |          | TODO                                        |
+| [DPTR](./configs/rec/dptr/)                   | [ACM MM 2024](https://dl.acm.org/doi/10.1145/3664647.3681390)                                    |          |          | [fd-zs](https://github.com/fd-zs)           |
+| VIPTR                                         | [ACM CIKM 2024](https://arxiv.org/abs/2401.10110)                                                |          |          | TODO                                        |
+| [IGTR](./configs/rec/igtr/)                   | [TPAMI 2025](https://ieeexplore.ieee.org/document/10820836)                                      | ✅       | ✅       |                                             |
+| [SMTR](./configs/rec/smtr/)                   | [AAAI 2025](https://ojs.aaai.org/index.php/AAAI/article/view/32285)                              | ✅       | ✅       |                                             |
+| [CPPD](./configs/rec/cppd/)                   | [TPAMI 2025](https://ieeexplore.ieee.org/document/10902187)                                      | ✅       | ✅       |                                             |
+| [FocalSVTR-CTC](./configs/rec/svtrs/)         | [AAAI 2025](https://ojs.aaai.org/index.php/AAAI/article/view/32285)                              | ✅       | ✅       |                                             |
+| [SVTRv2](./configs/rec/svtrv2/)               | [ICCV 2025](https://arxiv.org/abs/2411.15858)                                                    | ✅       | ✅       |                                             |
+| [ResNet+Trans-CTC](./configs/rec/svtrs/)      |                                                                                                  | ✅       | ✅       |                                             |
+| [ViT-CTC](./configs/rec/svtrs/)               |                                                                                                  | ✅       | ✅       |                                             |
+
+#### 核心贡献者
+
+______________________________________________________________________
+
+复旦大学[FVL实验室](https://fvl.fudan.edu.cn)的Yiming Lei ([pretto0](https://github.com/pretto0)), Xingsong Ye ([YesianRohn](https://github.com/YesianRohn)), and Shuai Zhao ([fd-zs](https://github.com/fd-zs))在Zhineng Chen老师([个人主页](https://zhinchenfd.github.io/))指导下完成了主要算法复现工作，感谢他们的贡献。
+
+### 场景文本检测(STD)
+
+开发中
+
+### 端到端文本识别(Text Spotting)
+
+开发中
+
+______________________________________________________________________
+
+## 引用
+
+如果我们的工作对您的研究有所帮助，请引用：
+
+```bibtex
+@inproceedings{Du2024SVTRv2,
+      title={SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition},
+      author={Yongkun Du and Zhineng Chen and Hongtao Xie and Caiyan Jia and Yu-Gang Jiang},
+      booktitle={ICCV},
+      year={2025}
+}
+```
+
+## 致谢
+
+本代码库基于[PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR)、[PytorchOCR](https://github.com/WenmuZhou/PytorchOCR)和[MMOCR](https://github.com/open-mmlab/mmocr)构建，感谢他们的出色工作！

__init__.py

@@ -0,0 +1,11 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+import os
+import sys
+
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.insert(0, os.path.abspath(os.path.join(__dir__, '..')))
+
+from tools.infer_e2e import OpenOCR, OpenDetector, OpenRecognizer

app.py

@@ -0,0 +1,117 @@
+import gradio as gr
+import torch
+from threading import Thread
+
+import numpy as np
+from openrec.postprocess.unirec_postprocess import clean_special_tokens
+from openrec.preprocess import create_operators, transform
+from tools.engine.config import Config
+from tools.utils.ckpt import load_ckpt
+from tools.infer_rec import build_rec_process
+
+
+def set_device(device):
+    if device == 'gpu' and torch.cuda.is_available():
+        device = torch.device(f'cuda:0')
+    else:
+        device = torch.device('cpu')
+    return device
+
+
+cfg = Config('configs/rec/unirec/focalsvtr_ardecoder_unirec.yml')
+cfg = cfg.cfg
+global_config = cfg['Global']
+
+from openrec.modeling.transformers_modeling.modeling_unirec import UniRecForConditionalGenerationNew
+from openrec.modeling.transformers_modeling.configuration_unirec import UniRecConfig
+from transformers import AutoTokenizer, TextIteratorStreamer
+
+tokenizer = AutoTokenizer.from_pretrained(global_config['vlm_ocr_config'])
+cfg_model = UniRecConfig.from_pretrained(global_config['vlm_ocr_config'])
+# cfg_model._attn_implementation = "flash_attention_2"
+cfg_model._attn_implementation = 'eager'
+
+model = UniRecForConditionalGenerationNew(config=cfg_model)
+load_ckpt(model, cfg)
+device = set_device(cfg['Global']['device'])
+model.eval()
+model.to(device=device)
+
+transforms, ratio_resize_flag = build_rec_process(cfg)
+ops = create_operators(transforms, global_config)
+
+
+# --- 2. Streaming generation function ---
+def stream_chat_with_image(input_image, history):
+    if input_image is None:
+        yield history + [('🖼️(empty)', 'Please upload an image first.')]
+        return
+
+    # Create TextIteratorStreamer
+    streamer = TextIteratorStreamer(
+        tokenizer,
+        skip_prompt=True,
+        skip_special_tokens=False
+    )
+
+    data = {'image': input_image}
+    batch = transform(data, ops[1:])
+    images = np.expand_dims(batch[0], axis=0)
+    images = torch.from_numpy(images).to(device=device)
+    inputs = {
+        'pixel_values': images,
+        'input_ids': None,
+        'attention_mask': None
+    }
+    generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=1024)
+    # Running generation in background thread
+    thread = Thread(target=model.generate, kwargs=generation_kwargs)
+    thread.start()
+
+    # Stream output
+    generated_text = ''
+    history = history + [('🖼️(image)', '')]
+    for new_text in streamer:
+        new_text = clean_special_tokens(new_text)
+        generated_text += new_text
+        history[-1] = ('🖼️(image)', generated_text)
+        yield history
+
+
+# --- 3. Gradio UI ---
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.HTML("""
+            <h1 style='text-align: center;'>
+                <a href="https://github.com/Topdu/OpenOCR">
+                    UniRec-0.1B: Unified Text and Formula Recognition with 0.1B Parameters
+                </a>
+            </h1>
+            <p style='text-align: center;'>
+               A ultralight unified text and formula recognition model 
+                (Created by <a href="https://fvl.fudan.edu.cn">FVL Lab</a>, 
+                <a href="https://github.com/Topdu/OpenOCR">OCR Team</a>)
+            </p>
+            <p style='text-align: center;'>
+                <a href="https://github.com/Topdu/OpenOCR/blob/main/docs/unirec.md">[Local GPU Deployment]</a>
+                for fast recognition experience
+            </p>"""
+            )
+    gr.Markdown('Upload an image, and the system will automatically recognize text and formulas.')
+    with gr.Row():
+        with gr.Column(scale=1):  # Left column: image + clear button
+            image_input = gr.Image(label='Upload Image or Paste Screenshot', type='pil')
+            clear = gr.ClearButton([image_input], value='Clear')
+        with gr.Column(scale=2):
+            chatbot = gr.Chatbot(
+                label='Result (Use LaTeX renderer to display formulas)',
+                show_copy_button=True,
+                height='auto'
+            )
+    clear.add([chatbot])
+
+    # Trigger after upload
+    image_input.upload(stream_chat_with_image, [image_input, chatbot], chatbot)
+
+# --- 4. Launch app ---
+if __name__ == '__main__':
+    demo.queue().launch(share=True)

configs/dataset/rec/evaluation.yaml

@@ -0,0 +1,41 @@
+root: ../evaluation
+task: str
+download_links:
+  # IC15_1811
+  - https://drive.usercontent.google.com/download?id=1eGY0kXNV1qVxeUpoGzs-ioUO-ky7msH6&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1BWv7aLoLAT7avY326gXP3GJF48UZpuBC&authuser=0&confirm=t
+  # SVT
+  - https://drive.usercontent.google.com/download?id=1ecEZ4cJ7dIbTCZRltE0s5KzUotQWagH-&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1OygBP7i9R-3Pwi6WodCcW31J8CUMugOJ&authuser=0&confirm=t
+  # IIIT5k
+  - https://drive.usercontent.google.com/download?id=1PJ9_IvIGZTS5hHdGLnpKuYKZcCO8jE0E&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=10P3MixSBt1v8k8_6aFfziC33Z5IlM6Uf&authuser=0&confirm=t
+  # IC13_857
+  - https://drive.usercontent.google.com/download?id=1-wMHOFBXJaOaY-UD00nDn6qw2s_8R4Vd&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1J1QCFtOFxFKiLJIgTqZ6eRo9Y5QGqHpA&authuser=0&confirm=t
+  # SVTP
+  - https://drive.usercontent.google.com/download?id=1kckwfZkdaHG8k_FW5IIJKUaYZkF21Hza&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1x61lm_ea7lvIdxNPMG-jy-5W0MxtdH0N&authuser=0&confirm=t
+  # CUTE80
+  - https://drive.usercontent.google.com/download?id=1Zv_91c81tinLy5Je89HPr-5wUSnqXKIB&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1OuJ6QoJ9AlyNHIM9j2WedAPxTnac7kyY&authuser=0&confirm=t
+filenames: 
+  # IC15_1811
+  - ../evaluation/IC15_1811/data.mdb
+  - ../evaluation/IC15_1811/lock.mdb
+  # SVT
+  - ../evaluation/SVT/data.mdb
+  - ../evaluation/SVT/lock.mdb
+  # IIIT5k
+  - ../evaluation/IIIT5k/data.mdb
+  - ../evaluation/IIIT5k/lock.mdb
+  # IC13_857
+  - ../evaluation/IC13_857/data.mdb
+  - ../evaluation/IC13_857/lock.mdb
+  # SVTP
+  - ../evaluation/SVTP/data.mdb
+  - ../evaluation/SVTP/lock.mdb
+  # CUTE80
+  - ../evaluation/CUTE80/data.mdb
+  - ../evaluation/CUTE80/lock.mdb
+check_validity: true

configs/dataset/rec/ltb.yaml

@@ -0,0 +1,9 @@
+root: ../ltb
+task: str
+download_links:
+  - https://drive.usercontent.google.com/download?id=16AEA1YGTsyVB44uEjKi4ZUV1snjCYBr4&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1xU4OStrOaI23bPG4flWAPWn2YrQe2bmY&authuser=0&confirm=t
+filenames:
+  - ../ltb/data.mdb
+  - ../ltb/lock.mdb
+check_validity: true

configs/dataset/rec/mjsynth.yaml

@@ -0,0 +1,11 @@
+root: ../synth
+task: str
+download_links:
+  - https://drive.usercontent.google.com/download?id=1FIoplSFZ-BKQoRDHDXsVMKa844e-K8PD&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1eckTvaeRtlTZvbO2orrVz-cIuIk6i87K&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1PBXTf-2PnmEvJBsqzJqxxRwzhAZGTiMG&authuser=0&confirm=t
+filenames:
+  - ../synth/MJ_train.zip
+  - ../synth/MJ_val.zip
+  - ../synth/MJ_test.zip
+check_validity: true

configs/dataset/rec/openvino.yaml

@@ -0,0 +1,25 @@
+root: ../OpenVINO
+task: str
+download_links:
+  # train_1
+  - https://drive.usercontent.google.com/download?id=1q23QAIRTyG0t-bBm4aAwRwiqB6VUfphw&authuser=0&confirm=
+  # train_2
+  - https://drive.usercontent.google.com/download?id=1AtbaJljM68cbZqi5lcM92d9VkQUCbSqI&authuser=0&confirm=
+  # train_5
+  - https://drive.usercontent.google.com/download?id=1dejstYnJ8_sESuO_uvwi__jT1B8gPxf3&authuser=0&confirm=t
+  # train_f
+  - https://drive.usercontent.google.com/download?id=1C4akchTc7-yi1OS_sJ3KP693UKcnecke&authuser=0&confirm=t
+  # validation
+  - https://drive.usercontent.google.com/download?id=17TRzSQhuK_juAxAv3KmX0y13pQP2cz6R&authuser=0&confirm=t
+filenames:
+  # train_1
+  - ../OpenVINO/train_1.zip
+  # train_2
+  - ../OpenVINO/train_2.zip
+  # train_5
+  - ../OpenVINO/train_5.zip
+  # train_f
+  - ../OpenVINO/train_f.zip
+  # validation
+  - ../OpenVINO/validation.zip
+check_validity: true

configs/dataset/rec/ost.yaml

@@ -0,0 +1,17 @@
+root: ../OST
+task: str
+download_links:
+  # OST heavy
+  - https://drive.usercontent.google.com/download?id=1RGpIFbD_SRlrzZFBoVF_LGvetNx1-5pg&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1Th4MfDf44k0EBpIqCLqVoGRu6G-FP1hq&authuser=0&confirm=t
+  # OST weak
+  - https://drive.usercontent.google.com/download?id=1z5CTDJucUnvALG12Q4UXk1DDKJDd8WJn&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1V17TTkX3sjpV7v0km_F2SDCK0tL3k_ls&authuser=0&confirm=t
+filenames:
+  # OST heavy
+  - ../OST/heavy/data.mdb
+  - ../OST/heavy/lock.mdb
+  # OST weak
+  - ../OST/weak/data.mdb
+  - ../OST/weak/lock.mdb
+check_validity: true

configs/dataset/rec/synthtext.yaml

@@ -0,0 +1,7 @@
+root: ../synth
+task: str
+download_links:
+  - https://drive.usercontent.google.com/download?id=1T-enqkq6_l2HqrsV3da_h0oJ7CUKu_oc&authuser=0&confirm=t
+filenames:
+  - ../synth/ST.zip
+check_validity: true

configs/dataset/rec/test.yaml

@@ -0,0 +1,77 @@
+root: ../test
+task: str
+download_links:
+  # IC13_857
+  - https://drive.usercontent.google.com/download?id=1PZSCbe6_DI8MlCqCRWXGT2PP92_frIXq&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1qkN7NDg0zUHxUiZHAeEatDTqlsgpFWp3&authuser=0&confirm=t
+  # IC15_2077
+  - https://drive.usercontent.google.com/download?id=1dFkY3DNbr-Mepn3TWBiA9COEJ63fGFcp&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1UvVwLNZ3tS1YdTBa8MulPzjeVezKaDro&authuser=0&confirm=t
+  # SVTP
+  - https://drive.usercontent.google.com/download?id=1aofeerilxJ7J3S7QxuCEXbmXTpz8Xshx&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1rJ1KoO4K_VUxEAUN_bMgBGzK8_JZAAno&authuser=0&confirm=t
+  # IIIT5k
+  - https://drive.usercontent.google.com/download?id=1XFO2M1Kbgwv3-iTNTmhQXAEjNmKYOeoT&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1stwK2hFsyaV7HHsEG9EYgnUQebNb2_nG&authuser=0&confirm=t
+  # COCOv1.4
+  - https://drive.usercontent.google.com/download?id=1Se2QSGS19xx7Gfy-SUdX9mlAOr2eYsfA&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1xvekFi389QfkH7yS0JIVV0QzjhUspjDv&authuser=0&confirm=t
+  # IC15_1811
+  - https://drive.usercontent.google.com/download?id=1pHsw8wrThD9EGEE6AusQLZozefSj4iyR&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1TXZ1qHuKAksaAlvd3qMv4IHKnN-IJW9a&authuser=0&confirm=t
+  # Uber
+  - https://drive.usercontent.google.com/download?id=1L2j6BZeLTGQ1FIl8HB_D3AFiWLltGV5r&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=12DUj28yzLWxFO_gfMfSjTkRujYD5MNEE&authuser=0&confirm=t
+  # IC13_1095
+  - https://drive.usercontent.google.com/download?id=1fu8onMt3Z6fDLNAiHcm-sQ2qCXduE-FU&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1OQAZtLj8U2Cl4L0ErGFsz6vGIVTTWasD&authuser=0&confirm=t
+  # IC13_1015
+  - https://drive.usercontent.google.com/download?id=1mbsfuvWB282HYfn9tbqcj1nUDkLXcSNB&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1QGogU_hV-oN7iY2POutdD2LDcmK6plnV&authuser=0&confirm=t
+  # ArT
+  - https://drive.usercontent.google.com/download?id=1-53knSy-uTSngCG7wyBngVyTuTCmdnWl&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=172EsSaf7BVaB1ORtohi-Jc_8SuUKZGGf&authuser=0&confirm=t
+  # SVT
+  - https://drive.usercontent.google.com/download?id=1p7aVUr9Yr7c4X4YUBvk2-YP28rraHjn9&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1ALmhvSleZ0yf-lcdbQPP3M9Zc3oqnXij&authuser=0&confirm=t
+  # CUTE80  
+  - https://drive.usercontent.google.com/download?id=1Ujr4axHKnu54P2rIGUhkjdM6XlhDYrI_&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1DvZi9L3MqjO2zRUyCg3YvP4qMAt2bsme&authuser=0&confirm=t
+filenames: 
+  # IC13_857
+  - ../test/IC13_857/data.mdb
+  - ../test/IC13_857/lock.mdb
+  # IC15_2077
+  - ../test/IC15_2077/data.mdb
+  - ../test/IC15_2077/lock.mdb
+  # SVTP
+  - ../test/SVTP/data.mdb
+  - ../test/SVTP/lock.mdb
+  # IIIT5k
+  - ../test/IIIT5k/data.mdb
+  - ../test/IIIT5k/lock.mdb
+  # COCOv1.4
+  - ../test/COCOv1.4/data.mdb
+  - ../test/COCOv1.4/lock.mdb
+  # IC15_1811
+  - ../test/IC15_1811/data.mdb
+  - ../test/IC15_1811/lock.mdb
+  # Uber
+  - ../test/Uber/data.mdb
+  - ../test/Uber/lock.mdb
+  # IC13_1095
+  - ../test/IC13_1095/data.mdb
+  - ../test/IC13_1095/lock.mdb
+  # IC13_1015
+  - ../test/IC13_1015/data.mdb
+  - ../test/IC13_1015/lock.mdb
+  # ArT
+  - ../test/ArT/data.mdb
+  - ../test/ArT/lock.mdb
+  # SVT
+  - ../test/SVT/data.mdb
+  - ../test/SVT/lock.mdb
+  # CUTE80
+  - ../test/CUTE80/data.mdb
+  - ../test/CUTE80/lock.mdb
+check_validity: true

configs/dataset/rec/textocr.yaml

@@ -0,0 +1,13 @@
+root: ../TextOCR
+task: str
+download_links:
+  # train
+  - https://drive.usercontent.google.com/download?id=1jVjJFno4pnsU0Cp_kn4MIXQrChmELy92&authuser=0&confirm=
+  # val
+  - https://drive.usercontent.google.com/download?id=1ubIRu01MXIek6OvInu-XjaIbw6277-vw&authuser=0&confirm=t
+filenames:
+  # train
+  - ../TextOCR/train.zip
+  # val
+  - ../TextOCR/val.zip
+check_validity: true

configs/dataset/rec/textocr_horizontal.yaml

@@ -0,0 +1,13 @@
+root: ../TextOCR_horizontal
+task: str
+download_links:
+  # train
+  - https://drive.usercontent.google.com/download?id=1sWH6J11xbjQb8SH7fdG_8mIKVI81ZQy5&authuser=0&confirm=
+  # val
+  - https://drive.usercontent.google.com/download?id=1gIE-AU2o-5hvg288-bjphO6UkI5AEQ2d&authuser=0&confirm=t
+filenames:
+  # train
+  - ../TextOCR_horizontal/train.zip
+  # val
+  - ../TextOCR_horizontal/val.zip
+check_validity: true

configs/dataset/rec/union14m_b.yaml

@@ -0,0 +1,47 @@
+root: ../u14m
+task: str
+download_links:
+  # artistic
+  - https://drive.usercontent.google.com/download?id=1Je2DTuFHnkXDI99yDnm9Anl5naWaCQwd&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1xtT_Q0juBJUIvAG55qBxoVNNTECd2usZ&authuser=0&confirm=t
+  # contextless
+  - https://drive.usercontent.google.com/download?id=1_0OzyzWhZOmGrHkayFTVrzhrQrNRDRPR&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1PPgC42y3xoM9bR0HQFbDYbcT3PzMdD_y&authuser=0&confirm=t
+  # salient
+  - https://drive.usercontent.google.com/download?id=1tHLMYBmTqRnxvFOTT3dfLfQiundqFWfd&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=13NQgpAtCK0kh9M5E2pAUmKKEp6Qu5Xwj&authuser=0&confirm=t
+  # multi_words
+  - https://drive.usercontent.google.com/download?id=1IlnDKX3V_Vp9gsDGFB0xoqsVLH1vtxUI&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1mFFjC7C0CwevvkwFU9YeVbZBdps_3Qpb&authuser=0&confirm=t
+  # curve
+  - https://drive.usercontent.google.com/download?id=1MxhMd85cmhUtI2lmtXhZQuFk7lav0_fw&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1N03g-4e-kJG2mRvlM0c5TrwWAkd-iG-Q&authuser=0&confirm=t
+  # general
+  - https://drive.usercontent.google.com/download?id=1Oqt7OaycP466NWoDmoJ3FqS8YP3YRgvu&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1K0MrX5eYNt8IIGFHXCwg0_oI5OF5PPFO&authuser=0&confirm=t
+  # multi_oriented
+  - https://drive.usercontent.google.com/download?id=1TKZFcZPVk0ThqfF-AGhJk_OCLg0ykKbv&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1PAoLMUWuR7O2-7XRoKkNzQcSiznErQzD&authuser=0&confirm=t
+filenames:
+  # artistic
+  - ../u14m/artistic/data.mdb
+  - ../u14m/artistic/lock.mdb
+  # contextless
+  - ../u14m/contextless/data.mdb
+  - ../u14m/contextless/lock.mdb
+  # salient
+  - ../u14m/salient/data.mdb
+  - ../u14m/salient/lock.mdb
+  # multi_words
+  - ../u14m/multi_words/data.mdb
+  - ../u14m/multi_words/lock.mdb
+  # curve
+  - ../u14m/curve/data.mdb
+  - ../u14m/curve/lock.mdb
+  # general
+  - ../u14m/general/data.mdb
+  - ../u14m/general/lock.mdb
+  # multi_oriented
+  - ../u14m/multi_oriented/data.mdb
+  - ../u14m/multi_oriented/lock.mdb
+check_validity: true

configs/dataset/rec/union14m_l_filtered.yaml

@@ -0,0 +1,35 @@
+root: ../Union14M-L-LMDB-Filtered
+task: str
+download_links:
+  # train_challenging
+  - https://drive.usercontent.google.com/download?id=1etwzBgGHjsFsb0sygsaRnKbanW2PMe07&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1ly6FJfPjItwGlVQ-ifTrzzM3rVu3Ezhr&authuser=0&confirm=t
+  # train_easy
+  - https://drive.usercontent.google.com/download?id=1_zeNluTnywIaa5h3PN-Ah9tKyByypot7&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1caYLeQHDidXgVBDi9IWXbO1gg__DYq9a&authuser=0&confirm=t
+  # train_hard
+  - https://drive.usercontent.google.com/download?id=1eP6s2xyYPZX9gykvWA4VSOc3Fqul_UB_&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1-ZlCvocX8P5uVRclUXp_5DNGLDzd16EO&authuser=0&confirm=t
+  # train_medium
+  - https://drive.usercontent.google.com/download?id=1s_CoaLNJEr-UxHYiqZ5jOcliMCFiRUUy&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1Wpj6WVpZ5Ily77kVwfQ18CiZBzkgmEnF&authuser=0&confirm=t
+  # train_normal
+  - https://drive.usercontent.google.com/download?id=1jPt44arlAswl9cXZjzmVcdpptdTPpJ3I&authuser=0&confirm=t
+  - https://drive.usercontent.google.com/download?id=1Rfc5kE03AzOUv7B_eYcBhUV8KMQ2MZ1m&authuser=0&confirm=t
+filenames:
+  # train_challenging
+  - ../Union14M-L-LMDB-Filtered/train_challenging/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_challenging/lock.mdb
+  # train_easy
+  - ../Union14M-L-LMDB-Filtered/train_easy/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_easy/lock.mdb
+  # train_hard
+  - ../Union14M-L-LMDB-Filtered/train_hard/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_hard/lock.mdb
+  # train_medium
+  - ../Union14M-L-LMDB-Filtered/train_medium/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_medium/lock.mdb
+  # train_normal
+  - ../Union14M-L-LMDB-Filtered/train_normal/data.mdb
+  - ../Union14M-L-LMDB-Filtered/train_normal/lock.mdb
+check_validity: true

configs/det/dbnet/repvit_db.yml

@@ -0,0 +1,171 @@
+Global:
+  device: gpu
+  epoch_num: &epoch_num 500
+  log_smooth_window: 20
+  print_batch_step: 100
+  output_dir: ./output/det_repsvtr_db
+  save_epoch_step: [400, 25]
+  eval_batch_step:
+  - 0
+  - 1000
+  cal_metric_during_train: false
+  checkpoints:
+  pretrained_model: openocr_det_repvit_ch.pth
+  save_inference_dir: null
+  use_tensorboard: false
+  infer_img:
+  save_res_path: ./checkpoints/det_db/predicts_db.txt
+  distributed: true
+  model_type: det
+
+Architecture:
+  algorithm: DB_mobile
+  Backbone:
+    name: RepSVTR_det
+  Neck:
+    name: RSEFPN
+    out_channels: 96
+    shortcut: True
+  Head:
+    name: DBHead
+    k: 50
+
+Loss:
+  name: DBLoss
+  balance_loss: true
+  main_loss_type: DiceLoss
+  alpha: 5
+  beta: 10
+  ohem_ratio: 3
+
+Optimizer:
+  name: Adam
+  lr: 0.001
+  weight_decay: 5.0e-05
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: CosineAnnealingLR
+  warmup_epoch: 2
+
+PostProcess:
+  name: DBPostProcess
+  thresh: 0.3
+  box_thresh: 0.6
+  max_candidates: 1000
+  unclip_ratio: 1.5
+  score_mode: 'slow'
+
+Metric:
+  name: DetMetric
+  main_indicator: hmean
+
+Train:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ../icdar2015/text_localization/
+    label_file_list:
+      - ../icdar2015/text_localization/train_icdar2015_label.txt
+    ratio_list: [1.0]
+    transforms:
+    - DecodeImage:
+        img_mode: BGR
+        channel_first: false
+    - DetLabelEncode: null
+    - CopyPaste: null
+    - IaaAugment:
+        augmenter_args:
+        - type: Fliplr
+          args:
+            p: 0.5
+        - type: Affine
+          args:
+            rotate:
+            - -10
+            - 10
+        - type: Resize
+          args:
+            size:
+            - 0.5
+            - 3
+    - EastRandomCropData:
+        size:
+        - 640
+        - 640
+        max_tries: 50
+        keep_ratio: true
+    - MakeBorderMap:
+        shrink_ratio: 0.4
+        thresh_min: 0.3
+        thresh_max: 0.7
+        total_epoch: *epoch_num
+    - MakeShrinkMap:
+        shrink_ratio: 0.4
+        min_text_size: 8
+        total_epoch: *epoch_num
+    - NormalizeImage:
+        scale: 1./255.
+        mean:
+        - 0.485
+        - 0.456
+        - 0.406
+        std:
+        - 0.229
+        - 0.224
+        - 0.225
+        order: hwc
+    - ToCHWImage: null
+    - KeepKeys:
+        keep_keys:
+        - image
+        - threshold_map
+        - threshold_mask
+        - shrink_map
+        - shrink_mask
+  loader:
+    shuffle: true
+    drop_last: false
+    batch_size_per_card: 8
+    num_workers: 8
+
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ../icdar2015/text_localization/
+    label_file_list:
+      - ../icdar2015/text_localization/test_icdar2015_label.txt
+    transforms:
+    - DecodeImage:
+        img_mode: BGR
+        channel_first: false
+    - DetLabelEncode: null
+    - DetResizeForTest:
+        # image_shape: [1280, 1280]
+        # keep_ratio: True
+        # padding: True
+        limit_side_len: 960
+        limit_type: max
+    - NormalizeImage:
+        scale: 1./255.
+        mean:
+        - 0.485
+        - 0.456
+        - 0.406
+        std:
+        - 0.229
+        - 0.224
+        - 0.225
+        order: hwc
+    - ToCHWImage: null
+    - KeepKeys:
+        keep_keys:
+        - image
+        - shape
+        - polys
+        - ignore_tags
+  loader:
+    shuffle: false
+    drop_last: false
+    batch_size_per_card: 1
+    num_workers: 2
+profiler_options: null

configs/rec/abinet/resnet45_trans_abinet_lang.yml

@@ -0,0 +1,94 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/resnet45_trans_abinet_lang/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  # ./openocr_nolang_abinet_lang.pth
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_resnet45_trans_abinet_lang.txt
+  grad_clip_val: 20
+  use_amp: True
+
+Optimizer:
+  name: Adam
+  lr: 0.000267
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: MultiStepLR
+  milestones: [12]
+  gamma: 0.1
+
+Architecture:
+  model_type: rec
+  algorithm: ABINet
+  Transform:
+  Encoder:
+    name: ResNet45
+    in_channels: 3
+    strides: [2, 1, 2, 1, 1]
+  Decoder:
+    name: ABINetDecoder
+    iter_size: 3
+
+Loss:
+  name: ABINetLoss
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/abinet/resnet45_trans_abinet_wo_lang.yml

@@ -0,0 +1,93 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/resnet45_trans_abinet_wo_lang/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_resnet45_trans_abinet_wo_lang.txt
+  grad_clip_val: 20
+  use_amp: True
+
+Optimizer:
+  name: Adam
+  lr: 0.000267
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: MultiStepLR
+  milestones: [12]
+  gamma: 0.1
+
+Architecture:
+  model_type: rec
+  algorithm: ABINet
+  Transform:
+  Encoder:
+    name: ResNet45
+    in_channels: 3
+    strides: [2, 1, 2, 1, 1]
+  Decoder:
+    name: ABINetDecoder
+    iter_size: 0
+
+Loss:
+  name: ABINetLoss
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/abinet/svtrv2_abinet_lang.yml

@@ -0,0 +1,130 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_abinet_lang/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  # ./openocr_svtrv2_nolang_abinet_lang.pth
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_abinet_lang.txt
+  use_amp: True
+  grad_clip_val: 20
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: ABINet
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: ABINetDecoder
+    iter_size: 3
+    num_layers: 0
+
+Loss:
+  name: ABINetLoss
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/abinet/svtrv2_abinet_wo_lang.yml

@@ -0,0 +1,128 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_abinet_wo_lang/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_abinet_wo_lang.txt
+  use_amp: True
+  grad_clip_val: 20
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: ABINet
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: ABINetDecoder
+    iter_size: 0
+    num_layers: 0
+Loss:
+  name: ABINetLoss
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/aster/resnet31_lstm_aster_tps_on.yml

@@ -0,0 +1,93 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/resnet31_lstm_aster_tps_on
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/predicts_aster_tps.txt
+  use_amp: True
+  grad_clip_val: 1.0
+
+Optimizer:
+  name: Adam
+  lr: 0.002 # for 1gpus bs1024/gpu
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: aster
+  Transform:
+    name: Aster_TPS
+    tps_inputsize: [32, 64]
+    tps_outputsize: [32, 128]
+  Encoder:
+    name: ResNet_ASTER
+  Decoder:
+    name: ASTERDecoder
+
+Loss:
+  name: ARLoss
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 1024
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/aster/svtrv2_aster.yml

@@ -0,0 +1,127 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_aster
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_aster.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: aster
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: False
+  Decoder:
+    name: ASTERDecoder
+
+Loss:
+  name: ARLoss
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode: # Class handling label
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/aster/svtrv2_aster_tps_on.yml

@@ -0,0 +1,102 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_aster_tps_on
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_aster_tps_on.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: aster
+  Transform:
+    name: Aster_TPS
+    tps_inputsize: [32, 64]
+    tps_outputsize: [32, 128]
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: False
+  Decoder:
+    name: ASTERDecoder
+
+Loss:
+  name: ARLoss
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/autostr/autostr_lstm_aster_tps_on.yml

@@ -0,0 +1,95 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/autostr_lstm_aster_tps_on
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_autostr_lstm_aster_tps_on.txt
+  use_amp: True
+  grad_clip_val: 1.0
+
+Optimizer:
+  name: Adam
+  lr: 0.002 # for 4gpus bs256/gpu
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: autostr
+  Transform:
+    name: Aster_TPS
+    tps_inputsize: [32, 64]
+    tps_outputsize: [32, 128]
+  Encoder:
+    name: AutoSTREncoder
+    stride_stages: '[(2, 2), (2, 1), (2, 2), (2, 1), (2, 1)]'
+    conv_op_ids: [2, 2, 2, 2, 2, 2, 2, 1, 2, 2, 4, 1, 1, 6, 6]
+  Decoder:
+    name: ASTERDecoder
+
+Loss:
+  name: ARLoss
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/busnet/svtrv2_busnet.yml

@@ -0,0 +1,135 @@
+Global:
+  device: gpu
+  epoch_num: 10
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_busnet/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  # ./output/rec/u14m_filter/svtrv2_busnet_pretraining/best.pth
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_busnet.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: BUSBet
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: False
+  Decoder:
+    name: BUSDecoder
+    nhead: 6
+    num_layers: 6
+    dim_feedforward: 1536
+    ignore_index: &ignore_index 100
+    pretraining: False
+    # return_id: 2
+Loss:
+  name: ABINetLoss
+  ignore_index: *ignore_index
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/busnet/svtrv2_busnet_pretraining.yml

@@ -0,0 +1,134 @@
+Global:
+  device: gpu
+  epoch_num: 10
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_busnet_pretraining/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_busnet_pretraining.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: BUSBet
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: False
+  Decoder:
+    name: BUSDecoder
+    nhead: 6
+    num_layers: 6
+    dim_feedforward: 1536
+    ignore_index: &ignore_index 100
+    pretraining: True
+    # return_id: 0
+Loss:
+  name: ABINetLoss
+  ignore_index: *ignore_index
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/busnet/vit_busnet.yml

@@ -0,0 +1,104 @@
+Global:
+  device: gpu
+  epoch_num: 10
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/vit_busnet/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_vit_busnet.txt
+  grad_clip_val: 20
+  use_amp: True
+
+Optimizer:
+  name: Adam
+  lr: 0.00053 # 4gpus bs256/gpu
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: MultiStepLR
+  milestones: [6]
+  gamma: 0.1
+
+Architecture:
+  model_type: rec
+  algorithm: BUSBet
+  Transform:
+  Encoder:
+    name: ViT
+    img_size: [32,128]
+    patch_size: [4, 8]
+    embed_dim: 384
+    depth: 12
+    num_heads: 6
+    mlp_ratio: 4
+    qkv_bias: True
+  Decoder:
+    name: BUSDecoder
+    nhead: 6
+    num_layers: 6
+    dim_feedforward: 1536
+    ignore_index: &ignore_index 100
+    pretraining: False
+Loss:
+  name: ABINetLoss
+  ignore_index: *ignore_index
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/busnet/vit_busnet_pretraining.yml

@@ -0,0 +1,104 @@
+Global:
+  device: gpu
+  epoch_num: 10
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/vit_busnet_pretraining/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_vit_busnet_pretraining.txt
+  grad_clip_val: 20
+  use_amp: True
+
+Optimizer:
+  name: Adam
+  lr: 0.00053 # 4gpus bs256/gpu
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: MultiStepLR
+  milestones: [6]
+  gamma: 0.1
+
+Architecture:
+  model_type: rec
+  algorithm: BUSBet
+  Transform:
+  Encoder:
+    name: ViT
+    img_size: [32,128]
+    patch_size: [4, 8]
+    embed_dim: 384
+    depth: 12
+    num_heads: 6
+    mlp_ratio: 4
+    qkv_bias: True
+  Decoder:
+    name: BUSDecoder
+    nhead: 6
+    num_layers: 6
+    dim_feedforward: 1536
+    ignore_index: &ignore_index 100
+    pretraining: True
+Loss:
+  name: ABINetLoss
+  ignore_index: *ignore_index
+
+PostProcess:
+  name: ABINetLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ABINetLabelEncode:
+          ignore_index: *ignore_index
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/cam/convnextv2_cam_tps_on.yml

@@ -0,0 +1,118 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/convnextv2_cam_tps_on
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: False
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: &max_text_length 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_convnextv2_cam_tps_on.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.0008 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+  eps: 1.e-8
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 :  1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CAM
+  Transform:
+    name: Aster_TPS
+    tps_inputsize: [32, 64]
+    tps_outputsize: &img_shape [32, 128]
+  Encoder:
+    name: CAMEncoder
+    encoder_config:
+      name: ConvNeXtV2
+      depths: [2, 2, 8, 2]
+      dims: [80, 160, 320, 640]
+      strides: [[4,4], [2,1], [2,1], [1,1]]
+      drop_path_rate: 0.2
+      feat2d: True
+    nb_classes: 97
+    strides: [[4,4], [2,1], [2,1], [1,1]]
+    deform_stride: 2
+    stage_idx: 2
+    use_depthwise_unet: True
+    use_more_unet: False
+    binary_loss_type: BanlanceMultiClassCrossEntropyLoss
+    mid_size: True
+    d_embedding: 384
+  Decoder:
+    name: CAMDecoder
+    num_encoder_layers: -1
+    beam_size: 0
+    num_decoder_layers: 2
+    nhead: 8
+    max_len: *max_text_length
+
+Loss:
+  name: CAMLoss
+  loss_weight_binary: 1.5
+  label_smoothing: 0.
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CAMLabelEncode: # Class handling label
+          font_path: ./arial.ttf
+          image_shape: *img_shape
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'binary_mask'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/cam/convnextv2_tiny_cam_tps_on.yml

@@ -0,0 +1,118 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/convnextv2_tiny_cam_tps_on
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: False
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: &max_text_length 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_convnextv2_cam_tps_on.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.0008 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+  eps: 1.e-8
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 :  1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CAM
+  Transform:
+    name: Aster_TPS
+    tps_inputsize: [32, 64]
+    tps_outputsize: &img_shape [32, 128]
+  Encoder:
+    name: CAMEncoder
+    encoder_config:
+      name: ConvNeXtV2
+      depths: [3, 3, 9, 3]
+      dims: [96, 192, 384, 768]
+      strides: [[4,4], [2,1], [2,1], [1,1]]
+      drop_path_rate: 0.2
+      feat2d: True
+    nb_classes: 97
+    strides: [[4,4], [2,1], [2,1], [1,1]]
+    deform_stride: 2
+    stage_idx: 2
+    use_depthwise_unet: True
+    use_more_unet: False
+    binary_loss_type: BanlanceMultiClassCrossEntropyLoss
+    mid_size: False
+    d_embedding: 512
+  Decoder:
+    name: CAMDecoder
+    num_encoder_layers: -1
+    beam_size: 0
+    num_decoder_layers: 2
+    nhead: 8
+    max_len: *max_text_length
+
+Loss:
+  name: CAMLoss
+  loss_weight_binary: 1.5
+  label_smoothing: 0.
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CAMLabelEncode: # Class handling label
+          font_path: ./arial.ttf
+          image_shape: *img_shape
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'binary_mask'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/cam/svtrv2_cam_tps_on.yml

@@ -0,0 +1,123 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_cam_tps_on
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: False
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: &max_text_length 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_cam_tps_on.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 :  1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CAM
+  Transform:
+    name: Aster_TPS
+    tps_inputsize: [32, 64]
+    tps_outputsize: &img_shape [32, 128]
+  Encoder:
+    name: CAMEncoder
+    encoder_config:
+      name: SVTRv2LNConvTwo33
+      use_pos_embed: False
+      dims: [128, 256, 384]
+      depths: [6, 6, 6]
+      num_heads: [4, 8, 12]
+      mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+      local_k: [[5, 5], [5, 5], [-1, -1]]
+      sub_k: [[1, 1], [2, 1], [-1, -1]]
+      last_stage: false
+      feat2d: True
+    nb_classes: 97
+    strides: [[4, 4], [1, 1], [2, 1], [1, 1]]
+    k_size: [[2, 2], [1, 1], [2, 1], [1, 1]]
+    q_size: [4, 32]
+    deform_stride: 2
+    stage_idx: 2
+    use_depthwise_unet: True
+    use_more_unet: False
+    binary_loss_type: BanlanceMultiClassCrossEntropyLoss
+    mid_size: True
+    d_embedding: 384
+  Decoder:
+    name: CAMDecoder
+    num_encoder_layers: -1
+    beam_size: 0
+    num_decoder_layers: 2
+    nhead: 8
+    max_len: *max_text_length
+
+Loss:
+  name: CAMLoss
+  loss_weight_binary: 1.5
+  label_smoothing: 0.
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+PostProcess:
+  name: ARLabelDecode
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CAMLabelEncode: # Class handling label
+          font_path: ./arial.ttf
+          image_shape: *img_shape
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'binary_mask'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/cdistnet/resnet45_trans_cdistnet.yml

@@ -0,0 +1,93 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/resnet45_trans_cdistnet
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_resnet45_trans_cdistnet.txt
+  use_amp: True
+  grad_clip_val: 5
+
+Optimizer:
+  name: Adam
+  lr: 0.002 # for 4gpus bs256/gpu
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CDistNet
+  Transform:
+  Encoder:
+    name: ResNet45
+    in_channels: 3
+    strides: [2, 1, 2, 1, 1]
+  Decoder:
+    name: CDistNetDecoder
+    add_conv: True
+
+Loss:
+  name: ARLoss
+
+PostProcess:
+  name: ARLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/cdistnet/svtrv2_cdistnet.yml

@@ -0,0 +1,139 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_cdistnet/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_cdistnet.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 #4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CDistNet
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: CDistNetDecoder
+    add_conv: False
+    num_encoder_blocks: 0
+
+Loss:
+  name: ARLoss
+
+PostProcess:
+  name: ARLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/cppd/svtr_base_cppd.yml

@@ -0,0 +1,123 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtr_base_cppd/
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path
+  # ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtr_base_cppd.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CPPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRNet
+    img_size: [32, 128]
+    out_char_num: 25
+    out_channels: 256
+    patch_merging: 'Conv'
+    embed_dim: [128, 256, 384]
+    depth: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: ['Conv','Conv','Conv','Conv','Conv','Conv', 'Conv','Conv', 'Global','Global','Global','Global','Global','Global','Global','Global','Global','Global']
+    local_mixer: [[5, 5], [5, 5], [5, 5]]
+    last_stage: False
+    prenorm: True
+  Decoder:
+    name: CPPDDecoder
+    vis_seq: 64
+    num_layer: 2
+    pos_len: False
+    rec_layer: 1
+
+
+Loss:
+  name: CPPDLoss
+  ignore_index: 100
+  smoothing: True
+  pos_len: False
+  sideloss_weight: 1.0
+
+PostProcess:
+  name: CPPDLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation/
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 128
+    num_workers: 4

configs/rec/cppd/svtr_base_cppd_ch.yml

@@ -0,0 +1,126 @@
+Global:
+  device: gpu
+  epoch_num: 100
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/ch/svtr_base_cppd/
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 2000]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: False
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/ppocr_keys_v1.txt
+  # ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/ch/predicts_svtr_base_cppd.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.0005 # for 4gpus bs128/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: CosineAnnealingLR
+  warmup_epoch: 5
+
+Architecture:
+  model_type: rec
+  algorithm: CPPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRNet
+    img_size: [32, 256]
+    patch_merging: 'Conv'
+    embed_dim: [128, 256, 384]
+    depth: [6, 6, 4]
+    num_heads: [4, 8, 12]
+    mixer: ['Conv','Conv','Conv','Conv','Conv','Conv', 'Conv','Conv', 'Global','Global','Global','Global','Global','Global','Global','Global','Global','Global']
+    local_mixer: [[5, 5], [5, 5], [5, 5]]
+    last_stage: False
+    prenorm: True
+  Decoder:
+    name: CPPDDecoder
+    vis_seq: 128
+    num_layer: 3
+    pos_len: False
+    rec_layer: 1
+    ch: True
+
+
+Loss:
+  name: CPPDLoss
+  ignore_index: 7000
+  smoothing: True
+  pos_len: False
+  sideloss_weight: 1.0
+
+PostProcess:
+  name: CPPDLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../benchmark_bctr/benchmark_bctr_train
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          ch: True
+          ignore_index: 7000
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - SVTRResize:
+          image_shape: [3, 32, 256]
+          padding: True
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'label_index', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 128
+    drop_last: True
+    num_workers: 8
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../benchmark_bctr/benchmark_bctr_test/scene_test
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          ch: True
+          ignore_index: 7000
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - SVTRResize:
+          image_shape: [3, 32, 256]
+          padding: True
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 4

configs/rec/cppd/svtr_base_cppd_h8.yml

@@ -0,0 +1,123 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtr_base_h8_cppd/
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtr_base_cppd.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CPPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRNet
+    img_size: [32, 128]
+    out_char_num: 25
+    out_channels: 256
+    patch_merging: 'Conv'
+    embed_dim: [128, 256, 384]
+    depth: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    sub_k: [[1, 1], [2, 1]]
+    mixer: ['Conv','Conv','Conv','Conv','Conv','Conv', 'Conv','Conv', 'Global','Global','Global','Global','Global','Global','Global','Global','Global','Global']
+    local_mixer: [[5, 5], [5, 5], [5, 5]]
+    last_stage: False
+    prenorm: True
+  Decoder:
+    name: CPPDDecoder
+    vis_seq: 128
+    num_layer: 2
+    pos_len: False
+    rec_layer: 1
+
+Loss:
+  name: CPPDLoss
+  ignore_index: 100
+  smoothing: True
+  pos_len: False
+  sideloss_weight: 1.0
+
+PostProcess:
+  name: CPPDLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation/
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 128
+    num_workers: 4

configs/rec/cppd/svtr_base_cppd_syn.yml

@@ -0,0 +1,124 @@
+Global:
+  device: gpu
+  epoch_num: 60
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/syn/svtr_base_cppd/
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path
+  # ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/syn/predicts_svtr_base_cppd.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.0005 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: CosineAnnealingLR
+  warmup_epoch: 6
+
+Architecture:
+  model_type: rec
+  algorithm: CPPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRNet
+    img_size: [32, 100]
+    out_char_num: 25
+    out_channels: 256
+    patch_merging: 'Conv'
+    embed_dim: [128, 256, 384]
+    depth: [6, 6, 4]
+    num_heads: [4, 8, 12]
+    mixer: ['Conv','Conv','Conv','Conv','Conv','Conv', 'Conv','Conv', 'Global','Global','Global','Global','Global','Global','Global','Global','Global','Global']
+    local_mixer: [[5, 5], [5, 5], [5, 5]]
+    last_stage: False
+    prenorm: True
+  Decoder:
+    name: CPPDDecoder
+    vis_seq: 50
+    num_layer: 3
+    pos_len: False
+    rec_layer: 1
+
+
+Loss:
+  name: CPPDLoss
+  ignore_index: 100
+  smoothing: True
+  pos_len: False
+  sideloss_weight: 1.0
+
+PostProcess:
+  name: CPPDLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+
+Train:
+  dataset:
+    name: STRLMDBDataSet
+    data_dir: ./
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+    #   - SVTRRAug:
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - SVTRResize:
+          image_shape: [3, 32, 100]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 8
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation/
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - SVTRResize:
+          image_shape: [3, 32, 100]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 4

configs/rec/cppd/svtrv2_cppd.yml

@@ -0,0 +1,150 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_cppd/
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_cppd.txt
+  use_amp: True
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: CPPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: False
+  Decoder:
+    name: CPPDDecoder
+    ds: True
+    num_layer: 2
+    pos_len: False
+    rec_layer: 1
+
+
+Loss:
+  name: CPPDLoss
+  ignore_index: 100
+  smoothing: True
+  pos_len: False
+  sideloss_weight: 1.0
+
+PostProcess:
+  name: CPPDLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_node', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - CPPDLabelEncode: # Class handling label
+          pos_len: False
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/dan/resnet45_fpn_dan.yml

@@ -0,0 +1,98 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/resnet45_fpn_dan/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_resnet45_fpn_dan.txt
+  use_amp: True
+  grad_clip_val: 20
+
+Optimizer:
+  name: Adam
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.0
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: DAN
+  Transform:
+  Encoder:
+    name: ResNet45
+    in_channels: 3
+    strides: [2, 1, 2, 1, 1]
+    return_list: True
+  Decoder:
+    name: DANDecoder
+    max_len: 25
+    channels_list: [64, 128, 256, 512]
+    strides_list: [[2, 2], [1, 1], [1, 1]]
+    in_shape: [8, 32]
+    depth: 4
+
+Loss:
+  name: ARLoss
+
+PostProcess:
+  name: ARLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../Union14M-L-LMDB-Filtered
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode:
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ../evaluation
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode:
+      - RecTVResize:
+          image_shape: [32, 128]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/dan/svtrv2_dan.yml

@@ -0,0 +1,130 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/svtrv2_dan
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: ./tools/utils/EN_symbol_dict.txt
+  max_text_length: 25
+  use_space_char: False
+  save_res_path: ./output/rec/u14m_filter/predicts_svtrv2_dan.txt
+  use_amp: True
+  grad_clip_val: 20
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # 4gpus 256bs/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: DAN
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: DANDecoder
+    use_cam: False
+    max_len: 25
+
+Loss:
+  name: ARLoss
+
+PostProcess:
+  name: ARLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - ARLabelEncode:
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - ARLabelEncode:
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/dptr/dptr_parseq_pretrain.yml

@@ -0,0 +1,88 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: /share/ckpt/zhaoshuai/openocr/dptr_parseq/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  use_amp: True
+  save_res_path: /share/ckpt/zhaoshuai/openocr/dptr_parseq/predicts_dptr_parseq.txt
+  grad_clip_val: 20
+
+Optimizer:
+  name: AdamW
+  lr: 0.001485 # 2gpus 384bs/gpu
+  weight_decay: 0.
+  filter_bias_and_bn: False
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: DPTR
+  Decoder:
+    name: DptrParseq
+    decode_ar: True
+    refine_iters: 1
+    is_pretrain: True
+    ORP_path: /share/ckpt/zhaoshuai/parseq/clip_background.pth
+
+Loss:
+  name: PARSeqLoss
+
+PostProcess:
+  name: ARLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: TextLMDBDataSet
+    data_dir: /share/test/zhaoshuai/parseq-data/data/train/real/ArT
+    transforms:
+      - DPTRLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['clip_label', 'label'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: TextLMDBDataSet
+    data_dir: /share/test/zhaoshuai/parseq-data/data/val
+    transforms:
+      - DPTRLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['clip_label', 'label'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 256
+    num_workers: 2

configs/rec/focalsvtr/focalsvtr_ctc.yml

@@ -0,0 +1,137 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/u14m_filter/focalsvtr_ctc/
+  eval_epoch_step: [0, 1]
+  eval_batch_step: [0, 500]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path
+  # ./tools/utils/EN_symbol_dict.txt
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/u14m_filter/predicts_focalsvtr_ctc.txt
+
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065 # for 4gpus bs256/gpu
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: SVTR
+  Transform:
+  Encoder:
+    name: FocalSVTR
+    img_size: [32, 128]
+    depths: [6, 6, 6]
+    embed_dim: 96
+    sub_k: [[1, 1], [2, 1], [1, 1]]
+    focal_levels: [3, 3, 3]
+    out_channels: 256
+    last_stage: True
+  Decoder:
+    name: CTCDecoder
+
+Loss:
+  name: CTCLoss
+  zero_infinity: True
+
+PostProcess:
+  name: CTCLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+
+Train:
+  dataset:
+    name: RatioDataSet
+    ds_width: True
+    padding: &padding False
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - PARSeqAug:
+      - CTCLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length']
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: 12
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSet
+    ds_width: True
+    padding: True
+    data_dir_list: ['../evaluation/CUTE80',
+    '../evaluation/IC13_857',
+    '../evaluation/IC15_1811',
+    '../evaluation/IIIT5k',
+    '../evaluation/SVT',
+    '../evaluation/SVTP',
+    ]
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - CTCLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length']
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: 128
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 128
+    max_ratio: 12
+    num_workers: 4

configs/rec/gtc/svtrv2_lnconv_nrtr_gtc.yml

@@ -0,0 +1,168 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/svtrv2_lnconv_nrtr_gtc
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img: ../ltb/img
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/predicts_smtr.txt
+  use_amp: True
+  distributed: true
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: BGPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: GTCDecoder
+    infer_gtc: True
+    detach: False
+    gtc_decoder:
+      name: NRTRDecoder
+      num_encoder_layers: -1
+      beam_size: 0
+      num_decoder_layers: 2
+      nhead: 12
+      max_len: *max_text_length
+    ctc_decoder:
+      name: RCTCDecoder
+
+Loss:
+  name: GTCLoss
+  gtc_loss:
+    name: ARLoss
+
+PostProcess:
+  name: GTCLabelDecode
+  gtc_label_decode:
+    name: ARLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecGTCMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSet
+    ds_width: True
+    # max_ratio: &max_ratio 4
+    # min_ratio: 1
+    # base_shape: &base_shape [[64, 64], [96, 48], [112, 40], [128, 32]]
+    # base_h: &base_h 32
+    # padding: &padding False
+    padding: false
+    # padding_rand: true
+    # padding_doub: true
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - PARSeqAug:
+      - GTCLabelEncode: # Class handling label
+          gtc_label_encode:
+            name: ARLabelEncode
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'ctc_label', 'ctc_length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 4
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSet
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - GTCLabelEncode: # Class handling label
+          gtc_label_encode:
+            name: ARLabelEncode
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'ctc_label', 'ctc_length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/gtc/svtrv2_lnconv_smtr_gtc_long_infer.yml

@@ -0,0 +1,151 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/svtrv2_lnconv_smtr_gtc_long_infer
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 1000]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img: ../ltb/img
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/predicts_smtr.txt
+  use_amp: True
+  distributed: true
+
+Optimizer:
+  name: AdamW
+  lr: 0.000325
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: BGPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: GTCDecoder
+    infer_gtc: False
+    detach: False
+    gtc_decoder:
+      name: SMTRDecoder
+      num_layer: 1
+      ds: True
+      max_len: *max_text_length
+      next_mode: &next True
+      sub_str_len: &subsl 5
+    ctc_decoder:
+      name: RCTCDecoder
+
+Loss:
+  name: CTCLoss
+
+PostProcess:
+  name: CTCLabelDecode
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - CTCLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 128
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 12
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: False
+    data_dir_list: [
+      '../evaluation/CUTE80',
+      '../evaluation/IC13_857',
+      '../evaluation/IC15_1811',
+      '../evaluation/IIIT5k',
+      '../evaluation/SVT',
+      '../evaluation/SVTP',
+      ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - CTCLabelEncode: # Class handling label
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: *bs
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: False
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: *bs
+    max_ratio: *max_ratio
+    num_workers: 4

configs/rec/gtc/svtrv2_lnconv_smtr_gtc_smtr_long.yml

@@ -0,0 +1,150 @@
+Global:
+  device: gpu
+  epoch_num: 20
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/svtrv2_lnconv_smtr_gtc_nodetach_smtr_long_infer
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 1000]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/predicts_smtr.txt
+  use_amp: True
+  distributed: true
+
+Optimizer:
+  name: AdamW
+  lr: 0.000325
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 1.5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+Architecture:
+  model_type: rec
+  algorithm: BGPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: GTCDecoder
+    infer_gtc: True
+    detach: False
+    gtc_decoder:
+      name: SMTRDecoder
+      num_layer: 1
+      ds: True
+      max_len: *max_text_length
+      next_mode: &next True
+      sub_str_len: &subsl 5
+      infer_aug: True
+    ctc_decoder:
+      name: RCTCDecoder
+
+Loss:
+  name: GTCLoss
+  ctc_weight: 0.1
+  gtc_loss:
+    name: SMTRLoss
+
+PostProcess:
+  name: GTCLabelDecode
+  gtc_label_decode:
+    name: SMTRLabelDecode
+    next_mode: *next
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+  only_gtc: True
+
+Metric:
+  name: RecGTCMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - SMTRLabelEncode: # Class handling label
+          sub_str_len: *subsl
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_subs', 'label_next', 'length_subs',
+          'label_subs_pre', 'label_next_pre', 'length_subs_pre', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 12
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ../ltb/
+    label_file_list: ['../ltb/ultra_long_70_list.txt']
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - GTCLabelEncode: # Class handling label
+          gtc_label_encode:
+            name: ARLabelEncode
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: 200
+      - SliceResize:
+          image_shape: [3, 32, 128]
+          padding: False
+          max_ratio: 12
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'ctc_label', 'ctc_length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 1
+    num_workers: 2

configs/rec/gtc/svtrv2_lnconv_smtr_gtc_stream.yml

@@ -0,0 +1,152 @@
+Global:
+  device: gpu
+  epoch_num: 60
+  log_smooth_window: 20
+  print_batch_step: 10
+  output_dir: ./output/rec/svtrv2_lnconv_smtr_gtc_stream
+  save_epoch_step: [15, 1]
+  # evaluation is run every 2000 iterations
+  eval_batch_step: [0, 500]
+  eval_epoch_step: [0, 1]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  use_tensorboard: false
+  infer_img:
+  # for data or label process
+  character_dict_path: &character_dict_path ./tools/utils/EN_symbol_dict.txt # 96en
+  # ./tools/utils/ppocr_keys_v1.txt  # ch
+  max_text_length: &max_text_length 25
+  use_space_char: &use_space_char False
+  save_res_path: ./output/rec/predicts_smtr.txt
+  use_amp: True
+  distributed: true
+  grad_clip_val: 20
+
+Optimizer:
+  name: AdamW
+  lr: 0.00065
+  weight_decay: 0.05
+  filter_bias_and_bn: True
+
+LRScheduler:
+  name: OneCycleLR
+  warmup_epoch: 5 # pct_start 0.075*20 = 1.5ep
+  cycle_momentum: False
+
+Architecture:
+  model_type: rec
+  algorithm: BGPD
+  in_channels: 3
+  Transform:
+  Encoder:
+    name: SVTRv2LNConvTwo33
+    use_pos_embed: False
+    out_channels: 256
+    dims: [128, 256, 384]
+    depths: [6, 6, 6]
+    num_heads: [4, 8, 12]
+    mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','FGlobal','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
+    local_k: [[5, 5], [5, 5], [-1, -1]]
+    sub_k: [[1, 1], [2, 1], [-1, -1]]
+    last_stage: false
+    feat2d: True
+  Decoder:
+    name: GTCDecoder
+    infer_gtc: True
+    detach: False
+    gtc_decoder:
+      name: SMTRDecoder
+      num_layer: 1
+      ds: True
+      max_len: *max_text_length
+      next_mode: &next True
+      sub_str_len: &subsl 5
+      infer_aug: False
+    ctc_decoder:
+      name: RCTCDecoder
+
+Loss:
+  name: GTCLoss
+  ctc_weight: 0.25
+  gtc_loss:
+    name: SMTRLoss
+
+PostProcess:
+  name: GTCLabelDecode
+  gtc_label_decode:
+    name: SMTRLabelDecode
+    next_mode: *next
+  character_dict_path: *character_dict_path
+  use_space_char: *use_space_char
+  only_gtc: True
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+  stream: True
+
+Train:
+  dataset:
+    name: RatioDataSetTVResize
+    ds_width: True
+    padding: false
+    data_dir_list: ['../Union14M-L-LMDB-Filtered/filter_train_challenging',
+    '../Union14M-L-LMDB-Filtered/filter_train_hard',
+    '../Union14M-L-LMDB-Filtered/filter_train_medium',
+    '../Union14M-L-LMDB-Filtered/filter_train_normal',
+    '../Union14M-L-LMDB-Filtered/filter_train_easy',
+    ]
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - PARSeqAugPIL:
+      - SMTRLabelEncode: # Class handling label
+          sub_str_len: *subsl
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'label_subs', 'label_next', 'length_subs',
+          'label_subs_pre', 'label_next_pre', 'length_subs_pre', 'length'] # dataloader will return list in this order
+  sampler:
+    name: RatioSampler
+    scales: [[128, 32]] # w, h
+    # divide_factor: to ensure the width and height dimensions can be devided by downsampling multiple
+    first_bs: &bs 256
+    fix_bs: false
+    divided_factor: [4, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: True
+    batch_size_per_card: *bs
+    drop_last: True
+    max_ratio: &max_ratio 12
+    num_workers: 4
+
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ../ltb/
+    label_file_list: ['../ltb/ultra_long_70_list.txt']
+    transforms:
+      - DecodeImagePIL: # load image
+          img_mode: RGB
+      - GTCLabelEncode: # Class handling label
+          gtc_label_encode:
+            name: ARLabelEncode
+          character_dict_path: *character_dict_path
+          use_space_char: *use_space_char
+          max_text_length: *max_text_length
+      - SliceTVResize:
+          image_shape: [32, 128]
+          padding: False
+          max_ratio: 4
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'ctc_label', 'ctc_length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 1
+    num_workers: 2

configs/rec/igtr/readme.md

@@ -0,0 +1,192 @@
+# IGTR
+
+- [IGTR](#igtr)
+  - [1. Introduction](#1-introduction)
+  - [2. Environment](#2-environment)
+    - [Dataset Preparation](#dataset-preparation)
+  - [3. Model Training / Evaluation](#3-model-training--evaluation)
+  - [Citation](#citation)
+
+<a name="1"></a>
+
+## 1. Introduction
+
+Paper:
+
+> [Instruction-Guided Scene Text Recognition](https://arxiv.org/abs/2401.17851),
+> Yongkun Du, Zhineng Chen, Yuchen Su, Caiyan Jia, Yu-Gang Jiang,
+> TPAMI
+
+<a name="model"></a>
+Multi-modal models have shown appealing performance in visual recognition tasks, as free-form text-guided training evokes the ability to understand fine-grained visual content. However, current models cannot be trivially applied to scene text recognition (STR) due to the compositional difference between natural and text images. We propose a novel instruction-guided scene text recognition (IGTR) paradigm that formulates STR as an instruction learning problem and understands text images by predicting character attributes, e.g., character frequency, position, etc. IGTR first devises $\\left \\langle condition,question,answer\\right \\rangle$ instruction triplets, providing rich and diverse descriptions of character attributes. To effectively learn these attributes through question-answering, IGTR develops a lightweight instruction encoder, a cross-modal feature fusion module and a multi-task answer head, which guides nuanced text image understanding. Furthermore, IGTR realizes different recognition pipelines simply by using different instructions, enabling a character-understanding-based text reasoning paradigm that differs from current methods considerably. Experiments on English and Chinese benchmarks show that IGTR outperforms existing models by significant margins, while maintaining a small model size and fast inference speed. Moreover, by adjusting the sampling of instructions, IGTR offers an elegant way to tackle the recognition of rarely appearing and morphologically similar characters, which were previous challenges.
+
+<a name="model"></a>
+The accuracy (%) and model files of IGTR on the public dataset of scene text recognition are as follows:
+
+- Trained on Synth dataset(MJ+ST), test on Common Benchmarks, training and test datasets both from [PARSeq](https://github.com/baudm/parseq).
+
+|  Model  | IC13<br/>857 | SVT  | IIIT5k<br/>3000 | IC15<br/>1811 | SVTP | CUTE80 |  Avg  |                                        Config&Model&Log                                         |
+| :-----: | :----------: | :--: | :-------------: | :-----------: | :--: | :----: | :---: | :---------------------------------------------------------------------------------------------: |
+| IGTR-PD |     97.6     | 95.2 |      97.6       |     88.4      | 91.6 |  95.5  | 94.30 | [link](https://drive.google.com/drive/folders/1Pv0CW2hiWC_dIyaB74W1fsXqiX3z5yXA?usp=drive_link) |
+| IGTR-AR |     98.6     | 95.7 |      98.2       |     88.4      | 92.4 |  95.5  | 94.78 |                                            as above                                             |
+
+- Test on Union14M-L benchmark, from [Union14M](https://github.com/Mountchicken/Union14M/).
+
+|  Model  | Curve | Multi-<br/>Oriented | Artistic | Contextless | Salient | Multi-<br/>word | General |  Avg  |    Config&Model&Log     |
+| :-----: | :---: | :-----------------: | :------: | :---------: | :-----: | :-------------: | :-----: | :---: | :---------------------: |
+| IGTR-PD | 76.9  |        30.6         |   59.1   |    63.3     |  77.8   |      62.5       |  66.7   | 62.40 | Same as the above table |
+| IGTR-AR | 78.4  |        31.9         |   61.3   |    66.5     |  80.2   |      69.3       |  67.9   | 65.07 |        as above         |
+
+- Trained on Union14M-L training dataset.
+
+|    Model     | IC13<br/>857 | SVT  | IIIT5k<br/>3000 | IC15<br/>1811 | SVTP | CUTE80 |  Avg  |                                        Config&Model&Log                                         |
+| :----------: | :----------: | :--: | :-------------: | :-----------: | :--: | :----: | :---: | :---------------------------------------------------------------------------------------------: |
+|   IGTR-PD    |     97.7     | 97.7 |      98.3       |     89.8      | 93.7 |  97.9  | 95.86 | [link](https://drive.google.com/drive/folders/1ZGlzDqEzjrBg8qG2wBkbOm3bLRzFbTzo?usp=drive_link) |
+|   IGTR-AR    |     98.1     | 98.4 |      98.7       |     90.5      | 94.9 |  98.3  | 96.48 |                                            as above                                             |
+| IGTR-PD-60ep |     97.9     | 98.3 |      99.2       |     90.8      | 93.7 |  97.6  | 96.24 | [link](https://drive.google.com/drive/folders/1ik4hxZDRsjU1RbCA19nwE45Kg1bCnMoa?usp=drive_link) |
+| IGTR-AR-60ep |     98.4     | 98.1 |      99.3       |     91.5      | 94.3 |  97.6  | 96.54 |                                            as above                                             |
+|  IGTR-PD-PT  |     98.6     | 98.0 |      99.1       |     91.7      | 96.8 |  99.0  | 97.20 | [link](https://drive.google.com/drive/folders/1QM0EWV66IfYI1G0Xm066V2zJA62hH6-1?usp=drive_link) |
+|  IGTR-AR-PT  |     98.8     | 98.3 |      99.2       |     92.0      | 96.8 |  99.0  | 97.34 |                                            as above                                             |
+
+|    Model     | Curve | Multi-<br/>Oriented | Artistic | Contextless | Salient | Multi-<br/>word | General |  Avg  |    Config&Model&Log     |
+| :----------: | :---: | :-----------------: | :------: | :---------: | :-----: | :-------------: | :-----: | :---: | :---------------------: |
+|   IGTR-PD    | 88.1  |        89.9         |   74.2   |    80.3     |  82.8   |      79.2       |  83.0   | 82.51 | Same as the above table |
+|   IGTR-AR    | 90.4  |        91.2         |   77.0   |    82.4     |  84.7   |      84.0       |  84.4   | 84.86 |        as above         |
+| IGTR-PD-60ep | 90.0  |        92.1         |   77.5   |    82.8     |  86.0   |      83.0       |  84.8   | 85.18 | Same as the above table |
+| IGTR-AR-60ep | 91.0  |        93.0         |   78.7   |    84.6     |  87.3   |      84.8       |  85.6   | 86.43 |        as above         |
+|  IGTR-PD-PT  | 92.4  |        92.1         |   80.7   |    83.6     |  87.7   |      86.9       |  85.0   | 86.92 | Same as the above table |
+|  IGTR-AR-PT  | 93.0  |        92.9         |   81.3   |    83.4     |  88.6   |      88.7       |  85.6   | 87.65 |        as above         |
+
+- Trained and test on Chinese dataset, from [Chinese Benckmark](https://github.com/FudanVI/benchmarking-chinese-text-recognition).
+
+|    Model    | Scene | Web  | Document | Handwriting |  Avg  |                                        Config&Model&Log                                         |
+| :---------: | :---: | :--: | :------: | :---------: | :---: | :---------------------------------------------------------------------------------------------: |
+|   IGTR-PD   | 73.1  | 74.8 |   98.6   |    52.5     | 74.75 |                                                                                                 |
+|   IGTR-AR   | 75.1  | 76.4 |   98.7   |    55.3     | 76.37 |                                                                                                 |
+| IGTR-PD-TS  | 73.5  | 75.9 |   98.7   |    54.5     | 75.65 | [link](https://drive.google.com/drive/folders/1H3VRdGHjhawd6fkSC-qlBzVzvYYTpHRg?usp=drive_link) |
+| IGTR-AR-TS  | 75.6  | 77.0 |   98.8   |    57.3     | 77.17 |                                            as above                                             |
+| IGTR-PD-Aug | 79.5  | 80.0 |   99.4   |    58.9     | 79.45 | [link](https://drive.google.com/drive/folders/1XFQkCILwcFwA7iYyQY9crnrouaI5sqcZ?usp=drive_link) |
+| IGTR-AR-Aug | 82.0  | 81.7 |   99.5   |    63.8     | 81.74 |                                            as above                                             |
+
+Download all Configs, Models, and Logs from [Google Drive](https://drive.google.com/drive/folders/1mSRDg9Mj5R6PspAdFGXZHDHTCQmjkd8d?usp=drive_link).
+
+<a name="2"></a>
+
+## 2. Environment
+
+- [PyTorch](http://pytorch.org/) version >= 1.13.0
+- Python version >= 3.7
+
+```shell
+git clone -b develop https://github.com/Topdu/OpenOCR.git
+cd OpenOCR
+# A100 Ubuntu 20.04 Cuda 11.8
+conda create -n openocr python==3.8
+conda activate openocr
+conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=11.8 -c pytorch -c nvidia
+pip install -r requirements.txt
+```
+
+#### Dataset Preparation
+
+- [English dataset download](https://github.com/baudm/parseq)
+
+- [Union14M-L-LMDB-Filtered download](https://drive.google.com/drive/folders/1OlDWJZgvd6s4S09S3IGeAI90jI0i7AB_?usp=sharing)
+
+- [Chinese dataset download](https://github.com/fudanvi/benchmarking-chinese-text-recognition#download)
+
+The expected filesystem structure is as follows:
+
+```
+benchmark_bctr
+├── benchmark_bctr_test
+│   ├── document_test
+│   ├── handwriting_test
+│   ├── scene_test
+│   └── web_test
+└── benchmark_bctr_train
+    ├── document_train
+    ├── handwriting_train
+    ├── scene_train
+    └── web_train
+evaluation
+├── CUTE80
+├── IC13_857
+├── IC15_1811
+├── IIIT5k
+├── SVT
+└── SVTP
+OpenOCR
+synth
+├── MJ
+│   ├── test
+│   ├── train
+│   └── val
+└── ST
+test # from PARSeq
+├── ArT
+├── COCOv1.4
+├── CUTE80
+├── IC13_1015
+├── IC13_1095
+├── IC13_857
+├── IC15_1811
+├── IC15_2077
+├── IIIT5k
+├── SVT
+├── SVTP
+└── Uber
+u14m # lmdb format
+├── artistic
+├── contextless
+├── curve
+├── general
+├── multi_oriented
+├── multi_words
+└── salient
+Union14M-L-LMDB-Filtered # lmdb format
+├── train_challenging
+├── train_easy
+├── train_hard
+├── train_medium
+└── train_normal
+```
+
+<a name="3"></a>
+
+## 3. Model Training / Evaluation
+
+Training:
+
+```shell
+# The configuration file is available from the link provided in the table above.
+# Multi GPU training
+CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 tools/train_rec.py --c PATH/svtr_base_igtr_XXX.yml
+```
+
+Evaluation:
+
+```shell
+# The configuration file is available from the link provided in the table above.
+# en
+python tools/eval_rec_all_en.py --c PATH/svtr_base_igtr_syn.yml
+# ch
+python tools/eval_rec_all_ch.py --c PATH/svtr_base_igtr_ch_aug.yml
+```
+
+## Citation
+
+If you find our method useful for your reserach, please cite:
+
+```bibtex
+@article{Du2024IGTR,
+  title     = {Instruction-Guided Scene Text Recognition},
+  author    = {Du, Yongkun and Chen, Zhineng and Su, Yuchen and Jia, Caiyan and Jiang, Yu-Gang},
+  journal   = {CoRR},
+  eprinttype = {arXiv},
+  primaryClass={cs.CV},
+  volume    = {abs/2401.17851},
+  year      = {2024},
+  url       = {https://arxiv.org/abs/2401.17851}
+}
+```