SeqFace: Learning discriminative features by using face sequences
Abstract Deep convolutional neural networks (CNNs) have greatly improved the Face Recognition (FR) performance in recent years. Almost all CNNs in FR are trained on the carefully labeled datasets containing plenty of identities. However, such high‐quality datasets are very expensive to collect, whic...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2021-09-01
|
| Series: | IET Image Processing |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/ipr2.12243 |
| _version_ | 1811266533175853056 |
|---|---|
| author | Wei Hu Yangyu Huang Fan Zhang Ruirui Li Hengchao Li |
| author_facet | Wei Hu Yangyu Huang Fan Zhang Ruirui Li Hengchao Li |
| author_sort | Wei Hu |
| collection | DOAJ |
| description | Abstract Deep convolutional neural networks (CNNs) have greatly improved the Face Recognition (FR) performance in recent years. Almost all CNNs in FR are trained on the carefully labeled datasets containing plenty of identities. However, such high‐quality datasets are very expensive to collect, which restricts many researchers to achieve state‐of‐the‐art performance. In this paper, a framework, called SeqFace, for learning discriminative face features is proposed. Besides a traditional identity training dataset, the designed SeqFace can train CNNs by using an additional dataset which includes a large number of face sequences collected from videos. Moreover, the label smoothing regularization (LSR) and a new proposed discriminative sequence agent (DSA) loss are employed to enhance the discrimination power of deep face features via making full use of the sequence data. Only with a single ResNet model, the method achieves very competitive performance on several face recognition benchmarks, including LFW, YTF, CFP, AgeDB, and MegaFace. The code and model are publicly available at the website https://github.com/huangyangyu/SeqFace. |
| first_indexed | 2024-04-12T20:43:48Z |
| format | Article |
| id | doaj.art-5f179228b2c940f2a27f8773b6242e83 |
| institution | Directory Open Access Journal |
| issn | 1751-9659 1751-9667 |
| language | English |
| last_indexed | 2024-04-12T20:43:48Z |
| publishDate | 2021-09-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Image Processing |
| spelling | doaj.art-5f179228b2c940f2a27f8773b6242e832022-12-22T03:17:20ZengWileyIET Image Processing1751-96591751-96672021-09-0115112548255810.1049/ipr2.12243SeqFace: Learning discriminative features by using face sequencesWei Hu0Yangyu Huang1Fan Zhang2Ruirui Li3Hengchao Li4College of Information Science and Technology Beijing University of Chemical Technology Beijing ChinaMicrosoft Research Asia Beijing ChinaCollege of Information Science and Technology Beijing University of Chemical Technology Beijing ChinaCollege of Information Science and Technology Beijing University of Chemical Technology Beijing ChinaSichuan Provincial Key Laboratory of Information Coding and Transmission Southwest Jiaotong University Chengdu ChinaAbstract Deep convolutional neural networks (CNNs) have greatly improved the Face Recognition (FR) performance in recent years. Almost all CNNs in FR are trained on the carefully labeled datasets containing plenty of identities. However, such high‐quality datasets are very expensive to collect, which restricts many researchers to achieve state‐of‐the‐art performance. In this paper, a framework, called SeqFace, for learning discriminative face features is proposed. Besides a traditional identity training dataset, the designed SeqFace can train CNNs by using an additional dataset which includes a large number of face sequences collected from videos. Moreover, the label smoothing regularization (LSR) and a new proposed discriminative sequence agent (DSA) loss are employed to enhance the discrimination power of deep face features via making full use of the sequence data. Only with a single ResNet model, the method achieves very competitive performance on several face recognition benchmarks, including LFW, YTF, CFP, AgeDB, and MegaFace. The code and model are publicly available at the website https://github.com/huangyangyu/SeqFace.https://doi.org/10.1049/ipr2.12243Image recognitionComputer vision and image processing techniquesMachine learning (artificial intelligence)Neural netsCNNsface recognition |
| spellingShingle | Wei Hu Yangyu Huang Fan Zhang Ruirui Li Hengchao Li SeqFace: Learning discriminative features by using face sequences IET Image Processing Image recognition Computer vision and image processing techniques Machine learning (artificial intelligence) Neural nets CNNs face recognition |
| title | SeqFace: Learning discriminative features by using face sequences |
| title_full | SeqFace: Learning discriminative features by using face sequences |
| title_fullStr | SeqFace: Learning discriminative features by using face sequences |
| title_full_unstemmed | SeqFace: Learning discriminative features by using face sequences |
| title_short | SeqFace: Learning discriminative features by using face sequences |
| title_sort | seqface learning discriminative features by using face sequences |
| topic | Image recognition Computer vision and image processing techniques Machine learning (artificial intelligence) Neural nets CNNs face recognition |
| url | https://doi.org/10.1049/ipr2.12243 |
| work_keys_str_mv | AT weihu seqfacelearningdiscriminativefeaturesbyusingfacesequences AT yangyuhuang seqfacelearningdiscriminativefeaturesbyusingfacesequences AT fanzhang seqfacelearningdiscriminativefeaturesbyusingfacesequences AT ruiruili seqfacelearningdiscriminativefeaturesbyusingfacesequences AT hengchaoli seqfacelearningdiscriminativefeaturesbyusingfacesequences |