Killing two birds with one stone: efficient and robust training of face recognition CNNs by partial FC
Author(s)
Type
Conference Paper
Abstract
Learning discriminative deep feature embeddings by using million-scale in-the-wild datasets and margin-based softmax loss is the current state-of-the-art approach for face recognition. However, the memory and computing cost of the Fully Connected (FC) layer linearly scales up to the number of identities in the training set. Besides, the largescale training data inevitably suffers from inter-class conflict and long-tailed distribution. In this paper, we propose a sparsely updating variant of the FC layer, named Partial FC (PFC). In each iteration, positive class centers and a random subset of negative class centers are selected to compute the margin-based softmax loss. All class centers are still maintained throughout the whole training process, but only a subset is selected and updated in each iteration. Therefore, the computing requirement, the probability of inter-class conflict, and the frequency of passive update on tail class centers, are dramatically reduced. Extensive experiments across different training data and backbones (e.g. CNN and ViT) confirm the effectiveness, robustness and efficiency of the proposed PFC. The source code is available at https://github.com/deepinsight/insightface/tree/master/recognition.
Date Issued
2022-09-27
Date Acceptance
2022-06-18
Citation
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp.4032-4041
ISSN
1063-6919
Publisher
IEEE Computer Society
Start Page
4032
End Page
4041
Journal / Book Title
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Copyright Statement
© 2022 IEEE. This CVPR paper is the Open Access versions, provided by the Computer Vision Foundation. Except for the watermark, they are identical to the accepted versions; the final published version of the proceedings is available on IEEE Xplore.
Source
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Subjects
Computer Science
Computer Science, Artificial Intelligence
Imaging Science & Photographic Technology
Science & Technology
Technology
Publication Status
Published
Start Date
2022-06-18
Finish Date
2022-06-24
Coverage Spatial
LA, New Orleans