Maximum A Posteriori Estimation of Distances Between Deep Features in Still-to-Video Face Recognition

TL;DR

This paper introduces a novel statistical method for still-to-video face recognition that maximizes the joint likelihood of distances between deep features, improving accuracy over traditional methods.

Contribution

It presents a maximum a posteriori estimation approach for face recognition that leverages the asymptotic normality of Kullback-Leibler divergences, enhancing recognition accuracy.

Findings

Achieves 3-5% higher accuracy than conventional methods.

Validates approach on LFW, YTF, and IJB-A datasets.

Compatible with state-of-the-art deep features and dissimilarity measures.

Abstract

The paper deals with the still-to-video face recognition for the small sample size problem based on computation of distances between high-dimensional deep bottleneck features. We present the novel statistical recognition method, in which the still-to-video recognition task is casted into Maximum A Posteriori estimation. In this method we maximize the joint probabilistic density of the distances to all reference still images. It is shown that this likelihood can be estimated with the known asymptotically normal distribution of the Kullback-Leibler discriminations between nonnegative features. The experimental study with the LFW (Labeled Faces in the Wild), YTF (YouTube Faces) and IJB-A (IARPA Janus Benchmark A) datasets has been provided. We demonstrated, that the proposed approach can be applied with the state-of-the-art deep features and dissimilarity measures. Our algorithm achieves 3-5% higher accuracy when compared with conventional aggregation of decisions obtained for all frames.