Machine Learning - Face Detection & Recognition (E2E)

Overview

Pipeline: face detection (pretrained), embedding extraction, and identity matching via nearest neighbors in a vector index (e.g., FAISS).

Steps

1. Detect faces to get bounding boxes.
2. Crop/align faces and compute embeddings with a trained model.
3. Match embeddings against a gallery using cosine similarity or FAISS.

Embeddings + FAISS (snippet)

# pip install faiss-cpu numpy
import faiss, numpy as np

# gallery_embeddings: (N, D), query: (D,)
index = faiss.IndexFlatIP(D)  # cosine similarity if vectors are L2-normalized
index.add(gallery_embeddings.astype('float32'))
Dists, Idxs = index.search(query.reshape(1,-1).astype('float32'), k=5)
print(Idxs[0], Dists[0])