Computer Vision · intermediate · concept 106 of 176
Face Detection & Recognition
Finding faces in images, then matching them to identities via face embeddings, where verification is one-to-one and identification is one-to-many. Accuracy went superhuman in the deep-learning era, which is exactly why the ethics got harder: demographic error gaps, surveillance concerns, and outright bans in some jurisdictions are part of the topic, not a footnote.
Key terms
Face embeddingsVerification vs identificationLiveness detectionNIST FRVTDemographic error gaps
Learn these first
Where you meet it in the real world
Phone unlock, passport e-gates, photo library grouping, and the regulation debates around all three
Videos
▶ Detecting Faces (Viola Jones Algorithm) - Computerphile ↗
Computerphile · YouTube
Guides and articles
Courses, papers, and more