venthanvi/fracture-api

Add data quality notebook, model improvements, and HF root fix

Refactor: ml/ package structure, improved Grad-CAM, radiology UX features

Fix libgl1-mesa-glx → libgl1 for Debian trixie

Add OpenCV system libs + opencv-headless for grad-cam

Add Grad-CAM debug logging

Switch to python:3.11-slim + CPU torch to fix conda/pip numpy conflict

Pin numpy<2 to fix RuntimeError: Numpy is not available

Use pytorch base image to speed up HF Space build

Trigger HF Space rebuild

Add full traceback logging for Grad-CAM errors

Fix Grad-CAM target layer for tf_efficientnetv2_m

Add README with HF Spaces Docker metadata

Add Dockerfile for Hugging Face Spaces deployment

Fix Render startup timeout: lazy-load model on first request

Fix Render start command and sys.path for api imports

Prepare backend for Render deployment

Overhaul kaggle_train.ipynb: fix all bugs, add HF upload

Fix argparse crash in Jupyter/Kaggle (parse_known_args)

Fix PyTorch dtype error in kaggle_train.ipynb

Update kaggle_train.ipynb: git clone instead of code dataset

Prepare frontend for Vercel deployment

Add full project structure: ML pipeline, API, Frontend

Fix Kaggle path detection for images and code datasets