How jeffheaton/dynaface works
Project type and purpose
• Dynaface is a two-part system:
• A Python library (dynaface-lib) that implements the core facial-landmark detection, symmetry measures, data loading pipelines, and ML inference.
• Its purpose is to automatically locate 97 facial landmarks on still images or video, compute symmetry/asymmetry measures via advanced AI (CNN/GNN), and export results for research.
File structure
Top level
• dynaface-app: UI code, command definitions, worker threads, deploy subfolders (Mac/Windows) with PyInstaller specs, build scripts.
• dynaface-lib: core library packages (dynaface/…), a spiga submodule with data loaders, inference framework, CNN/GNN models, pretrained weights, plus tests and examples.
• CI/CD: Jenkinsfile-lib (build/test library), Jenkinsfile-deploy (build and package app), build_all.sh, scripts/test_whl.*.
• Configuration: requirements.txt (app), constraints.txt, setup.py/setup.cfg (lib), .flake8, .gitignore.
README for additional insights
• Confirms dual nature (app + library), AI references, output formats, user guide link.