[Chore/ML] Move Model Weights from Git to Hugging Face Model Hub

[jpdevhub]

Problem

The Models/species_mobilenetv3.pth file (6.0 MB) was committed directly to the Git repository using git add -f. While 6 MB is manageable today, as we add more models (freshness model, YOLO, etc.) the repository will become too large to clone.

Goal

Host all model weights on the Hugging Face Model Hub and download them at runtime, keeping Git for code only.

Acceptance Criteria

• Create a public model repo at https://huggingface.co/jpdevhub/freshscan-models
• Upload species_mobilenetv3.pth to the HF Model Hub
• Create backend/download_models.py script that uses huggingface_hub to download weights to a local Models/ directory if not already present
• backend/startup.sh updated to call python download_models.py before starting uvicorn
• Models/ directory added to .gitignore
• Models/species_mobilenetv3.pth removed from Git history using git filter-repo or git-lfs migrate (or BFG Repo Cleaner)
• CONTRIBUTING.md updated with a note: 'Never commit .pth / .onnx / .bin model weights to Git'
• README.md updated with model download instructions

Commands Reference

pip install huggingface_hub
python -c "from huggingface_hub import hf_hub_download; hf_hub_download('jpdevhub/freshscan-models', 'species_mobilenetv3.pth', local_dir='Models/')"

[arcgod-design]

Hi @jpdevhub! I'd love to work on this as part of SSoC 2026. I'm @arcgod-design.

My Approach:

• Create a public model repo at https://huggingface.co/jpdevhub/freshscan-models
• Upload species_mobilenetv3.pth to the HF Model Hub
• Create backend/download_models.py script using huggingface_hub to download weights to a local Models/ directory if not already present
• Update backend/Dockerfile and startup scripts to call download_models.py before the app starts
• Remove the large .pth file from Git history using git filter-repo
• Update .gitignore to exclude Models/*.pth

Could you please assign this issue to me? Ready to start immediately!