Abstract

We propose a method to push the frontiers of uncon-strained face recognition in the wild, focusing on the problem of extreme pose variations. As opposed to current tech-niques which either expect a single model to learn poseinvariance through massive amounts of training data, orwhich normalize images to a single frontal pose, our methodexplicitly tackles pose variation by using multiple posespecific models and rendered face images. We leverage deep Convolutional Neural Networks (CNNs) to learn discriminative representations we call Pose-Aware Models (PAMs) using 500K images from the CASIA WebFace dataset. We present a comparative evaluation on the new IARPA Janus Benchmark A (IJB-A) and PIPA datasets. On these datasets PAMs achieve remarkably better performance than commercial products and surprisingly also outperform methodsthat are specifically fine-tuned on the target dataset.