Unsupervised Learning of Face Representations

TL;DR

This paper introduces an unsupervised CNN training method for face recognition using video data, achieving high accuracy on benchmarks without manual labeling, especially effective for low-resolution faces typical in surveillance.

Contribution

It presents a novel unsupervised training approach that leverages video data to learn discriminative face representations without manual labels.

Findings

Achieves higher verification accuracy on LFW than hand-crafted features.

Surpasses state-of-the-art deep networks like VGG-Face on low-resolution faces.

Effective in realistic surveillance scenarios with small face images.

Abstract

We present an approach for unsupervised training of CNNs in order to learn discriminative face representations. We mine supervised training data by noting that multiple faces in the same video frame must belong to different persons and the same face tracked across multiple frames must belong to the same person. We obtain millions of face pairs from hundreds of videos without using any manual supervision. Although faces extracted from videos have a lower spatial resolution than those which are available as part of standard supervised face datasets such as LFW and CASIA-WebFace, the former represent a much more realistic setting, e.g. in surveillance scenarios where most of the faces detected are very small. We train our CNNs with the relatively low resolution faces extracted from video frames collected, and achieve a higher verification accuracy on the benchmark LFW dataset cf. hand-crafted features such as LBPs, and even surpasses the performance of state-of-the-art deep networks such as VGG-Face, when they are made to work with low resolution input images.