Face verification under real-world conditions: prototype bundle
Code and trained model for the Gradio prototype of the IE University AI/ML group project "Facial Recognition in Real-World Conditions: How Lighting, Occlusion and Ageing Affect Face Verification".
Contents: the app (app.py), the minimal package code it needs (src/face_verification/), and the trained baseline (results/models/baseline_fold0.joblib). No dataset images are included.
Model: HOG features → PCA (256) → RBF-kernel SVM, trained on LFW fold 0's training people (validation accuracy 74%, AUC 0.81). Face detection: MTCNN from facenet-pytorch (MIT).
Run it: use the team's Colab notebook (run_prototype_colab.ipynb), or locally:
pip install scikit-learn==1.9.1 gradio==6.31.0 torch torchvision scikit-image joblib matplotlib pandas
pip install --no-deps facenet-pytorch==2.6.0
python app.py
Research prototype only, not an identity-check system. Uploaded photos are used only for the comparison and are deleted from the server within an hour.