Face Recognition
Detect faces and generate 512-dimensional ArcFace embeddings for identity matching and verification
Why do anything?
Face detection and recognition require specialized models. Generic image embeddings don't capture facial identity.
Why now?
Identity verification, security, and content moderation need reliable face matching at scale.
Why this feature?
ArcFace embeddings (512D) optimized for face verification. Detects faces with bounding boxes and landmarks.
How It Works
Face identity extractor combines MTCNN detection with ArcFace embeddings for face recognition.
Detection
MTCNN detects faces with bounding boxes
Alignment
Facial landmark-based face alignment
Embedding
ArcFace generates 512D identity embeddings
Storage
Store in Qdrant with face metadata
Why This Approach
ArcFace provides state-of-the-art face verification accuracy. 512D embeddings capture identity effectively.
Where This Is Used
Integration
client.collections.create(feature_extractor={"feature_extractor_name": "face_identity_extractor", "version": "v1"})