README.md

metadata

title: RefDiffNet
emoji: 🔬
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 4.44.0
python_version: '3.10'
app_file: app.py
pinned: false
license: agpl-3.0

RefDiffNet — PCB Reference–Defect Enrichment (A11_CA)

Space: vinayedula/RefDiffNet

Gradio demo for the A11_CA prebackbone trained with YOLO12n on HR-IPCB.

• Inputs: defect PCB image + golden reference image
• Output: enriched image (enriched = defect + α · gate · delta)

Weights: weights/prebackbone_a11_ca.pt (~few MB, prebackbone only — not full best.pt)

Deploy to Hugging Face Spaces

1. Prepare the Space folder locally

From the VYOLO repo root:

cd /mnt/data/vinay/work/VYOLO
bash hf_prebackbone_demo/prepare_space.sh

This will:

• Copy the custom Ultralytics fork into hf_prebackbone_demo/vendor/
• Extract prebackbone_a11_ca.pt from best.pt (small file for the Space)
• Copy example image pairs into hf_prebackbone_demo/examples/

2. Create a new Space on Hugging Face

1. Go to huggingface.co/new-space
2. Choose Gradio SDK
3. Clone the empty Space repo locally

3. Push to vinayedula/RefDiffNet

cd hf_prebackbone_demo
bash deploy_to_refdiffnet.sh

Or manually:

git clone https://huggingface.co/spaces/vinayedula/RefDiffNet
cd RefDiffNet
# copy app.py, vendor/, weights/prebackbone_a11_ca.pt, examples/, etc.
git add . && git commit -m "RefDiffNet demo" && git push

Note: prebackbone_a11_ca.pt is only a few MB (no LFS needed). Full best.pt is optional.

4. Space settings (recommended)

Setting	Value
Hardware	CPU Basic (works) or GPU for faster loads
Secrets	Optional: HF_TOKEN if weights are in a private model repo

Alternative: host weights on the Hub

Upload prebackbone_a11_ca.pt to a model repo, then set:

HF_MODEL_REPO=YOUR_USER/YOUR_MODEL

The app downloads prebackbone_a11_ca.pt from that repo.

Run locally

cd hf_prebackbone_demo
pip install -r requirements.txt
pip install -e ../ultralytics   # custom fork with A11_CA
python extract_prebackbone_weights.py --ckpt /path/to/best.pt
export PREBACKBONE_ONLY_WEIGHTS=weights/prebackbone_a11_ca.pt
python app.py

Open http://localhost:7860

Environment variables

Variable	Description
PREBACKBONE_ONLY_WEIGHTS	Path to prebackbone_a11_ca.pt
PREBACKBONE_FULL_CKPT	Full best.pt (only for one-time extraction)
HF_MODEL_REPO	Hub repo with prebackbone_a11_ca.pt
PREBACKBONE_IMGSZ	Letterbox size (default 640)
ULTRALYTICS_ROOT	Path to ultralytics repo if not vendored
PORT	Gradio port (default 7860)