arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1807.11649cs.CV
收藏

人脸识别的魔鬼在噪音中

The Devil of Face Recognition is in the Noise

SenseTime Research(商汤科技研究院);University of California San Diego(加州大学圣迭戈分校);Nanyang Technological University(南洋理工大学)

Fei Wang, Liren Chen, Cheng Li, Shiyao Huang, Yanjie Chen, Chen Qian, Chen Change Loy

摘要:

人脸识别数据集规模的不断增长使我们能够训练强大的卷积网络进行人脸识别。虽然已经设计了各种架构和损失函数,但我们对现有数据集中固有的标签噪声的来源和后果仍然有有限的了解。我们做出了以下贡献:1)我们贡献了流行的人脸数据库的清洁子集,即,MegaFace和MS-Celeb-1 M数据集,并构建一个新的大规模噪声控制的IMDb-Face数据集。2)使用原始数据集和清洁的子集,我们分析和分析MegaFace和MS-Celeb-1 M的标签噪声特性。我们表明,需要几个订单更多的样本,以达到相同的精度产生一个干净的子集。3)我们研究不同类型的噪音之间的联系,即,标签翻转和离群值,具有人脸识别模型的准确性。4)我们研究了提高数据清洁度的方法,包括对数据标签策略对注释准确性的影响进行全面的用户研究。IMDb-Face数据集已在https://github.com/fwang91/IMDb-Face上发布。

英文摘要:

The growing scale of face recognition datasets empowers us to train strong convolutional networks for face recognition. While a variety of architectures and loss functions have been devised, we still have a limited understanding of the source and consequence of label noise inherent in existing datasets. We make the following contributions: 1) We contribute cleaned subsets of popular face databases, i.e., MegaFace and MS-Celeb-1M datasets, and build a new large-scale noise-controlled IMDb-Face dataset. 2) With the original datasets and cleaned subsets, we profile and analyze label noise properties of MegaFace and MS-Celeb-1M. We show that a few orders more samples are needed to achieve the same accuracy yielded by a clean subset. 3) We study the association between different types of noise, i.e., label flips and outliers, with the accuracy of face recognition models. 4) We investigate ways to improve data cleanliness, including a comprehensive user study on the influence of data labeling strategies to annotation accuracy. The IMDb-Face dataset has been released on https://github.com/fwang91/IMDb-Face.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑