Face Recognition

The question

Recognition depends on consistent preprocessing and a clear split between pretrained representations and the classifier trained for a particular dataset.

How it works

• Align detected faces before embedding extraction.
• Use reproducible per-identity train/test splits.
• Train downstream classifiers and serialize model artifacts for inference.

Evidence & results

• Public source contains the implemented alignment, embedding and classification workflow.

Scope & next steps

• Upstream FaceNet benchmark scores are not personally achieved project results.
• This learning project does not establish deployment suitability or performance across populations.