AttFC: Attention Fully-Connected Layer for Large-Scale Face Recognition with One GPU

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Zhuowen, Si, Yain-Whar, Yuan, Xiaochen, Duan, Junwei, Wang, Ke, Li, Xiaofan, Zhang, Xinyuan, Gong, Xueyuan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912267036000256
author Zheng, Zhuowen
Si, Yain-Whar
Yuan, Xiaochen
Duan, Junwei
Wang, Ke
Li, Xiaofan
Zhang, Xinyuan
Gong, Xueyuan
author_facet Zheng, Zhuowen
Si, Yain-Whar
Yuan, Xiaochen
Duan, Junwei
Wang, Ke
Li, Xiaofan
Zhang, Xinyuan
Gong, Xueyuan
contents Nowadays, with the advancement of deep neural networks (DNNs) and the availability of large-scale datasets, the face recognition (FR) model has achieved exceptional performance. However, since the parameter magnitude of the fully connected (FC) layer directly depends on the number of identities in the dataset. If training the FR model on large-scale datasets, the size of the model parameter will be excessively huge, leading to substantial demand for computational resources, such as time and memory. This paper proposes the attention fully connected (AttFC) layer, which could significantly reduce computational resources. AttFC employs an attention loader to generate the generative class center (GCC), and dynamically store the class center with Dynamic Class Container (DCC). DCC only stores a small subset of all class centers in FC, thus its parameter count is substantially less than the FC layer. Also, training face recognition models on large-scale datasets with one GPU often encounter out-of-memory (OOM) issues. AttFC overcomes this and achieves comparable performance to state-of-the-art methods.
format Preprint
id arxiv_https___arxiv_org_abs_2503_06839
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AttFC: Attention Fully-Connected Layer for Large-Scale Face Recognition with One GPU
Zheng, Zhuowen
Si, Yain-Whar
Yuan, Xiaochen
Duan, Junwei
Wang, Ke
Li, Xiaofan
Zhang, Xinyuan
Gong, Xueyuan
Computer Vision and Pattern Recognition
Artificial Intelligence
Nowadays, with the advancement of deep neural networks (DNNs) and the availability of large-scale datasets, the face recognition (FR) model has achieved exceptional performance. However, since the parameter magnitude of the fully connected (FC) layer directly depends on the number of identities in the dataset. If training the FR model on large-scale datasets, the size of the model parameter will be excessively huge, leading to substantial demand for computational resources, such as time and memory. This paper proposes the attention fully connected (AttFC) layer, which could significantly reduce computational resources. AttFC employs an attention loader to generate the generative class center (GCC), and dynamically store the class center with Dynamic Class Container (DCC). DCC only stores a small subset of all class centers in FC, thus its parameter count is substantially less than the FC layer. Also, training face recognition models on large-scale datasets with one GPU often encounter out-of-memory (OOM) issues. AttFC overcomes this and achieves comparable performance to state-of-the-art methods.
title AttFC: Attention Fully-Connected Layer for Large-Scale Face Recognition with One GPU
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2503.06839