Remove torch from requirements (installed separately as CPU-only in Dockerfile)

Use CPU-only PyTorch for smaller Docker image

Update app: auto-find model path, add spinner, improve landing page

Upload best_model.pth with huggingface_hub

Upload src/train.py with huggingface_hub

Upload src/app/app.py with huggingface_hub

Add Streamlit config: disable XSRF/CORS for HF Spaces iframe compatibility

Fix 403 on file upload: disable XSRF/CORS, set writable dirs, increase upload limit

Fix: replace deprecated libgl1-mesa-glx with libgl1 for Debian Trixie

Upload README.md

Upload src/app/app.py with huggingface_hub

Upload src/train.py with huggingface_hub

Upload src/pipeline/full_pipeline.py with huggingface_hub

Upload src/pipeline/visualization.py with huggingface_hub

Upload src/pipeline/decision_engine.py with huggingface_hub

Upload src/pipeline/crack_analysis.py with huggingface_hub

Upload src/pipeline/mask_cleaning.py with huggingface_hub

Upload src/pipeline/inference.py with huggingface_hub

Upload src/pipeline/module_segmentation.py with huggingface_hub

Upload src/pipeline/preprocessing.py with huggingface_hub

Upload src/datasets/el_dataset.py with huggingface_hub

Upload src/losses/combined_loss.py with huggingface_hub

Upload src/models/unet.py with huggingface_hub

Upload src/losses/__init__.py

Upload src/pipeline/__init__.py

Upload src/__init__.py

Upload src/models/__init__.py

Upload src/agents/__init__.py

Upload src/app/__init__.py

Upload src/datasets/__init__.py

Upload requirements.txt

Upload Dockerfile

initial commit