README.md

metadata

title: Image Captioning Model
emoji: 💻
colorFrom: pink
colorTo: purple
sdk: gradio
sdk_version: 6.0.2
app_file: app.py
pinned: false
license: mit
short_description: Image_Captioning_Model

Image Captioning Model - Hugging Face Deployment

This folder contains everything needed to deploy the Image Captioning Model on Hugging Face Spaces.

📦 Files

• app.py - Gradio application (rebuilds model architecture and loads weights)
• requirements.txt - Python dependencies
• model_weights.h5 - Model weights only (22 MB)
• tokenizer_data.json - Tokenizer vocabulary (JSON format)
• model_config.pkl - Model configuration

🚀 Quick Deploy to Hugging Face Spaces

1. Go to Hugging Face Spaces
2. Click "Create new Space"
3. Choose:
   • Name: image-captioning-model (or your choice)
   • SDK: Gradio
   • Visibility: Public or Private
4. Upload all files from this folder
5. Wait for build (5-10 minutes)
6. Done! 🎉

💡 How It Works

This deployment uses a different approach than typical model loading:

• Instead of: Loading the full H5 model (which has compatibility issues)
• We do: Rebuild the model architecture in code and load only the weights

This avoids all the Keras version compatibility issues!

🔧 Technical Details

• TensorFlow: 2.10.0
• Gradio: 3.50.2
• Model Architecture: CNN-RNN (InceptionV3 + LSTM)
• Weights File: 22 MB (much smaller than full model)
• Tokenizer: JSON format (no pickle compatibility issues)

✅ Advantages of This Approach

1. ✅ No Keras version conflicts
2. ✅ Smaller file size (22 MB vs 66 MB)
3. ✅ Works with any TensorFlow 2.x version
4. ✅ Easy to modify model architecture if needed
5. ✅ No pickle security concerns

📝 Notes

• The model architecture is defined in app.py
• Weights are loaded using model.load_weights()
• This is the recommended approach for deploying Keras models

---

Created by: Prabhar Kumar Singh GitHub: Prabhat9801/Image_Captioning_Model

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference