WebFace260M: A Benchmark for Million-Scale Deep Face Recognition

Abstract

Face benchmarks empower the research community to train and evaluate high-performance face recognition systems. In this paper, we contribute a new million-scale recognition benchmark, containing uncurated 4M identities/260M faces (WebFace260M) and cleaned 2M identities/42M faces (WebFace42M) training data, as well as an elaborately designed time-constrained evaluation protocol. Firstly, we collect 4M name lists and download 260M faces from the Internet. Then, a Cleaning Automatically utilizing Self-Training (CAST) pipeline is devised to purify the tremendous WebFace260M, which is efficient and scalable. To the best of our knowledge, the cleaned WebFace42M is the largest public face recognition training set and we expect to close the data gap between academia and industry. Referring to practical deployments, Face Recognition Under Inference Time conStraint (FRUITS) protocol and a new test set with rich attributes are constructed. Besides, we gather a large-scale masked face sub-set for biometrics assessment under COVID-19. For a comprehensive evaluation of face matchers, three recognition tasks are performed under standard, masked and unbiased settings, respectively. Equipped with this benchmark, we delve into million-scale face recognition problems. A distributed framework is developed to train face recognition models efficiently without tampering with the performance. Enabled by WebFace42M, we reduce 40% failure rate on the challenging IJB-C set and rank 3rd among 430 entries on NIST-FRVT. Even 10% data (WebFace4M) shows superior performance compared with the public training sets. Furthermore, comprehensive baselines are established under the FRUITS-100/500/1000 milliseconds protocols. The proposed benchmark shows enormous potential on standard, masked and unbiased face recognition scenarios. Our WebFace260M website is https://www.face-benchmark.org.

Comparison of Data Encodings and Machine Learning Architectures for User Identification on Arbitrary Motion Sequences

“…Training the model on much larger datasets before getting applied in production will certainly perform even better, while no training is necessary when just applying the model to new test users. This is similar to face recognition, whose models are used in many photo applications today and do not need any on-device re-training, since they were trained on millions of individuals' faces [45].…”

“…We also observe that the performance is not saturated for the maximally available training users for our dataset. This suggests that additional users during training can substantially improve current results, which is a common observation in related applications such as face recognition, where production-grade models are typically trained on millions of subjects [45]. This motivates the collection of larger datasets for the task of VR user identification.…”

OCFR 2022: Competition on Occluded Face Recognition From Synthetically Generated Structure-Aware Occlusions

“…Considering that AdaFace improves largely the FR performance on low quality academic datasets [27], the aim of the proposed methods is to investigate the performance of AdaFace on the challenging scenario presenting low quality occluded faces. The two proposed solutions (AdaFace4M and AdaFace12M) are trained using million-scale subsets of the cleaned version of the ultra-large-scale face benchmark consisting of 4Midentities/260M faces (Web-Face260M) [41]. AdaFace12M is trained on a larger subset of 12M of the curated WebFace42M.…”

A model study of teaching method reform of computer laboratory course integrating internet of things technology

“…In the field of education, the wide application of IoT technology has brought about a sea change in teaching and learning. By connecting various devices, sensors and networks, IoT offers many new possibilities for education [ 2 ]. Teachers are utilizing IoT technology to better understand their progress and level of understanding of their students.…”