Exploiting the Tail Data for Long-Tailed Face Recognition
Long-tailed distribution generally exists in large-scale face datasets, which poses challenges for learning discriminative feature in face recognition. Although a few works conduct preliminary research on this problem, the value of the tail data is still underestimated. This paper addresses the long...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2022-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9887937/ |
| _version_ | 1818061060413849600 |
|---|---|
| author | Song Guo Rujie Liu Mengjiao Wang Meng Zhang Shijie Nie Septiana Lina Narishige Abe |
| author_facet | Song Guo Rujie Liu Mengjiao Wang Meng Zhang Shijie Nie Septiana Lina Narishige Abe |
| author_sort | Song Guo |
| collection | DOAJ |
| description | Long-tailed distribution generally exists in large-scale face datasets, which poses challenges for learning discriminative feature in face recognition. Although a few works conduct preliminary research on this problem, the value of the tail data is still underestimated. This paper addresses the long-tailed problem from the perspective of maximally exploiting the tail data. We propose a Joint Alternating Training (JAT) framework to learn discriminative feature from both the long-tailed data and the tail data by using alternating training strategy. JAT consists of two branches: 1) the long-tailed data branch is adopted to learn the universal discrimination information from the whole long-tailed data with instance-balanced sampling. 2) the tail data branch is designed to exploit the discriminative information in the tail data with class-balanced sampling. To compensate the insufficient samples and lack of intra-class variations, we apply data augmentation (DA) to the tail data. We further propose margin-based mixup (MarginMix) for data augmentation, which can deal with the nonlinearity of margin-based softmax loss and stabilize the training process in mixup. Furthermore, we obtain the best combination of strategies (i.e., JAT+DA+ MarginMix) for long-tailed face recognition, which can maximally exploit the discriminative information in the tail data while retaining the universal discrimination learned from the long-tailed data. Extensive experiments on 8 face datasets demonstrate that our proposed methods and combination of strategies can effectively address the long-tailed problem in face recognition. |
| first_indexed | 2024-12-10T13:42:18Z |
| format | Article |
| id | doaj.art-4d6ecbf621b344348230953d92f16260 |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-12-10T13:42:18Z |
| publishDate | 2022-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-4d6ecbf621b344348230953d92f162602022-12-22T01:46:37ZengIEEEIEEE Access2169-35362022-01-0110979459795310.1109/ACCESS.2022.32060409887937Exploiting the Tail Data for Long-Tailed Face RecognitionSong Guo0https://orcid.org/0000-0002-4505-4059Rujie Liu1Mengjiao Wang2Meng Zhang3https://orcid.org/0000-0003-1465-9714Shijie Nie4Septiana Lina5Narishige Abe6Fujitsu Research and Development Center Company Ltd., Beijing, ChinaFujitsu Research and Development Center Company Ltd., Beijing, ChinaFujitsu Research and Development Center Company Ltd., Beijing, ChinaFujitsu Research and Development Center Company Ltd., Beijing, ChinaFujitsu Research and Development Center Company Ltd., Beijing, ChinaFujitsu Laboratories Ltd., Kawasaki, JapanFujitsu Laboratories Ltd., Kawasaki, JapanLong-tailed distribution generally exists in large-scale face datasets, which poses challenges for learning discriminative feature in face recognition. Although a few works conduct preliminary research on this problem, the value of the tail data is still underestimated. This paper addresses the long-tailed problem from the perspective of maximally exploiting the tail data. We propose a Joint Alternating Training (JAT) framework to learn discriminative feature from both the long-tailed data and the tail data by using alternating training strategy. JAT consists of two branches: 1) the long-tailed data branch is adopted to learn the universal discrimination information from the whole long-tailed data with instance-balanced sampling. 2) the tail data branch is designed to exploit the discriminative information in the tail data with class-balanced sampling. To compensate the insufficient samples and lack of intra-class variations, we apply data augmentation (DA) to the tail data. We further propose margin-based mixup (MarginMix) for data augmentation, which can deal with the nonlinearity of margin-based softmax loss and stabilize the training process in mixup. Furthermore, we obtain the best combination of strategies (i.e., JAT+DA+ MarginMix) for long-tailed face recognition, which can maximally exploit the discriminative information in the tail data while retaining the universal discrimination learned from the long-tailed data. Extensive experiments on 8 face datasets demonstrate that our proposed methods and combination of strategies can effectively address the long-tailed problem in face recognition.https://ieeexplore.ieee.org/document/9887937/Face recognitionconvolutional neural networklong-tailed distributionmargin softmax lossdata augmentation |
| spellingShingle | Song Guo Rujie Liu Mengjiao Wang Meng Zhang Shijie Nie Septiana Lina Narishige Abe Exploiting the Tail Data for Long-Tailed Face Recognition IEEE Access Face recognition convolutional neural network long-tailed distribution margin softmax loss data augmentation |
| title | Exploiting the Tail Data for Long-Tailed Face Recognition |
| title_full | Exploiting the Tail Data for Long-Tailed Face Recognition |
| title_fullStr | Exploiting the Tail Data for Long-Tailed Face Recognition |
| title_full_unstemmed | Exploiting the Tail Data for Long-Tailed Face Recognition |
| title_short | Exploiting the Tail Data for Long-Tailed Face Recognition |
| title_sort | exploiting the tail data for long tailed face recognition |
| topic | Face recognition convolutional neural network long-tailed distribution margin softmax loss data augmentation |
| url | https://ieeexplore.ieee.org/document/9887937/ |
| work_keys_str_mv | AT songguo exploitingthetaildataforlongtailedfacerecognition AT rujieliu exploitingthetaildataforlongtailedfacerecognition AT mengjiaowang exploitingthetaildataforlongtailedfacerecognition AT mengzhang exploitingthetaildataforlongtailedfacerecognition AT shijienie exploitingthetaildataforlongtailedfacerecognition AT septianalina exploitingthetaildataforlongtailedfacerecognition AT narishigeabe exploitingthetaildataforlongtailedfacerecognition |