admin管理员组

文章数量:1652183

论文阅读 [CVPR-2022] An Efficient Training Approach for Very Large Scale Face Recognition

一种高效的超大规模人脸识别训练方法

studyai

搜索论文: An Efficient Training Approach for Very Large Scale Face Recognitionv

http://www.studyai/search/whole-site/?q=An=Efficient+Training+Approach+for+Very+Large+Scale+Face+Recognition

摘要(Abstract)

Face recognition has achieved significant progress in deep learning era due to the ultra-large-scale and well-labeled datasets.

由于超大规模和良好标记的数据集，人脸识别在深度学习时代取得了重大进展。

However, training on the outsize datasets is time-consuming and takes up a lot of hardware resource.

然而，在超大规模的数据集上进行训练是非常耗时的，并且占用了大量的硬件资源。

Therefore, designing an efficient training approach is indispensable. The heavy computational and memory costs mainly result from the million-level dimensionality of the
fully connected (FC) layer.

因此，设计一种高效的训练方法是必不可少的。沉重的计算和内存成本主要来自全连接（FC）层的百万级维度。

To this end, we propose a novel training approach, termed Faster Face Classification (F2C), to alleviate time and cost without sacrificing the performance.

为此，我们提出了一种新的训练方法，称为快速人脸分类（F2C），以减轻时间和成本，同时不影响性能。

This method adopts Dynamic Class Pool (DCP) for storing and updating the identities’ features dynamically, which could be regarded as a substitute for the FC layer.

该方法采用动态类库（DCP）来存储和动态更新身份的特征，它可以被视为FC层的替代品。

DCP is efficiently time-saving and cost-saving, as its smaller size with the independence from the whole face identities together.

DCP可以有效地节省时间和成本，因为它的体积较小，与整个面部身份无关。

We further validate the proposed F2C method across several face benchmarks and private datasets, and display comparable results, meanwhile the speed is faster than state-of-the-art FC-based methods in terms of recognition accuracy and hardware costs.

我们在几个人脸基准和私人数据集上进一步验证了所提出的F2C方法，并显示了可比较的结果，同时在识别精度和硬件成本方面比最先进的基于FC的方法更快。

Moreover, our method is further improved by a well-designed dual data loader including indentity-based and instance-based loaders, which makes it more efficient for updating DCP parameters.

此外，我们的方法通过精心设计的双数据加载器（包括基于身份的加载器和基于实例的加载器）得到了进一步的改善，这使得它在更新DCP参数时更加高效。

本文标签： 论文EfficientTrainingCVPRRecognition

版权声明：本文标题：论文阅读 [CVPR-2022] An Efficient Training Approach for Very Large Scale Face Recognition 内容由热心网友自发贡献，该文观点仅代表作者本人， 转载请联系作者并注明出处：https://www.elefans.com/dianzi/1729578696a1207339.html， 本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如发现本站有涉嫌抄袭侵权/违法违规的内容，一经查实，本站将立刻删除。