A web app that segments the image of a pet animal into three regions:
- Foreground (the pet)
- Background
- Boundary
It uses a U-Net model trained on the Oxford-IIIT Pet Dataset and is deployed using Gradio.
Architecture
The U-Net architecture is a convolutional neural network originally developed for biomedical image segmentation. Its encoder-decoder structure with skip connections makes it excellent for pixel-level segmentation tasks.
Tech Stack
| Component | Technology |
|---|---|
| Model | U-Net (TensorFlow / Keras) |
| Dataset | Oxford-IIIT Pet Dataset |
| Frontend | Gradio |
| Deployment | HuggingFace Spaces |