We present a Neural Aggregation Network (NAN) for video face recognition. The network takes a face video or face image set of a person with a variable number of face images as its input, and produces a compact and fixed-dimension feature representation. The whole network is composed of two modules. The feature embedding module is a deep Convolutional Neural Network (CNN), which maps each face image into a feature vector. The aggregation module consists of two attention blocks driven by a memory storing all the extracted features. It adaptively aggregates the features to form a single feature inside the convex hull spanned by them. Due to the attention mechanism, the aggregation is invariant to the image order. We found that NAN learns to advocate high-quality face images while repelling low-quality ones such as blurred, occluded and improperly exposed faces. The experiments on IJB-A, YouTube Face, Celebrity-1000 video face recognition benchmarks show that it consistently outperforms standard aggregation methods and achieves state-of-the-art accuracies.
from cs.AI updates on arXiv.org http://ift.tt/1SX15Wb
via IFTTT
No comments:
Post a Comment