sunraku16/Ayodo4

Ayodo Face Type

Ayodo Face Type is a simple Hugging Face Spaces app for face type classification, not person identification. It accepts one face image from upload or webcam, checks that the photo is usable, and returns a best-match face type for soft authentication or biometric tokenization demos.

This app uses interpretable facial geometry inspired by the feature families discussed in IBM's Diversity in Faces work, then maps those measurements into a practical 12-label taxonomy:

• Oval
• Round
• Square
• Rectangle
• Oblong
• Diamond
• Heart
• Inverted Triangle
• Triangle
• Pear
• Tapered Oval
• Tapered Square

Files

• app.py: Gradio app entry point for Hugging Face Spaces
• face_type_pipeline.py: detection, quality checks, alignment, geometry features, and 12-class prediction
• models/: YuNet face detector and LBF landmark model required for inference
• requirements.txt: minimal Python dependencies

Local Run

python3 -m pip install -r requirements.txt
python3 app.py

Hugging Face Spaces

1. Create a new Gradio Space.
2. Upload this repository as-is.
3. Keep app.py at the repo root.
4. Spaces will install requirements.txt and launch the app automatically.

Notes

• The app is designed for one visible face at a time.
• It rejects blurry, distant, or strongly side-posed photos.
• It does not perform identity recognition, age prediction, gender prediction, or medical diagnosis.