Partial least squares for face hashing

Abstract

Face recognition has been an active research topic in recent years due to its numerous applications in surveillance, biometrics, human-computer interaction and social media. In this work, we focus on face identification, which consists in determining the identity of a face image given a gallery of known faces. Specifically, the goal in this work is to provide an approach for face identification scalable to galleries consisting of several subjects. The proposed approach is inspired by locality-sensitive hashing (LSH) and partial least squares (PLS) for face identification. We employ a combination of four feature descriptors, resulting in a 120,059-dimensional vector and providing significant improvement over single feature descriptors. Results show significant reduction in the number of subjects evaluated in the face identification (reduced to 0.3% of the gallery), providing speedup up to 233 times compared to the brute-force approach and 58 times compared to previous works in the literature.