shravya11/shoe-dataset

Shoe Type Classification, Detection, and Web-Harvested Dataset

This repository contains three shoe datasets:

1. Shoe Type Classification Dataset: An image classification dataset for classifying different types of shoes (4,794 images).
2. COCO Shoe Detection Dataset: An object detection and part segmentation dataset annotated in COCO format (545 images).
3. Harvested Footwear Dataset: Unannotated footwear images programmatically gathered from Google Search (2,368 images).

• Grand Total Images: 7,707

---

1. Shoe Type Classification Dataset (4,794 images)

• Dataset Structure:
  • train/: 4,494 images total:
    • ballet_flats/: 600 images
    • boats/: 600 images
    • boots/: 249 images
    • brogues/: 600 images
    • clogs/: 600 images
    • flip_flops/: 249 images
    • loafers/: 249 images
    • sandals/: 249 images
    • sneakers/: 849 images
    • soccer_shoes/: 249 images
  • validation/: 300 images total (50 images per class):
    • boots/: 50 images
    • flip_flops/: 50 images
    • loafers/: 50 images
    • sandals/: 50 images
    • sneakers/: 50 images
    • soccer_shoes/: 50 images
• Format: Image Classification folder structure (subfolders named after classes containing raw images)

---

2. COCO Shoe Detection Dataset (545 images)

• Dataset Structure:
  • shoes_coco/train/: 281 images and _annotations.coco.json
  • shoes_coco/validation/: 149 images and _annotations.coco.json
  • shoes_coco/test/: 115 images and _annotations.coco.json
• Format: COCO Object Detection format (images and bounding box/polygon annotations in a single JSON file per split).
• Categories/Classes:
  • shoes (ID 0)
  • ankle (ID 1)
  • heel (ID 2)
  • left_foot (ID 3)
  • pinky (ID 4)
  • right_foot (ID 5)
  • toe (ID 6)

---

3. Harvested Footwear Dataset (2,368 images)

• Dataset Structure:
  • harvested_shoes/: 2,368 images total.
• Format: A single flat directory containing raw unannotated web-harvested images.
• Source: Gathered from Google Search.

---

License

CC BY 4.0