LCReg: long-tailed image classification with latent categories based recognition
In this work, we tackle the challenging problem of long-tailed image recognition. Previous long-tailed recognition approaches mainly focus on data augmentation or re-balancing strategies for the tail classes to give them more attention during model training. However, these methods are limited by the...
| Main Authors: | , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/170932 |
| _version_ | 1811685396716716032 |
|---|---|
| author | Liu, Weide Wu, Zhonghua Wang, Yiming Ding, Henghui Liu, Fayao Lin, Jie Lin, Guosheng |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Liu, Weide Wu, Zhonghua Wang, Yiming Ding, Henghui Liu, Fayao Lin, Jie Lin, Guosheng |
| author_sort | Liu, Weide |
| collection | NTU |
| description | In this work, we tackle the challenging problem of long-tailed image recognition. Previous long-tailed recognition approaches mainly focus on data augmentation or re-balancing strategies for the tail classes to give them more attention during model training. However, these methods are limited by the small number of training images for the tail classes, which results in poor feature representations. To address this issue, we propose the Latent Categories based long-tail Recognition (LCReg) method. Our hypothesis is that common latent features shared by head and tail classes can be used to improve feature representation. Specifically, we learn a set of class-agnostic latent features shared by both head and tail classes, and then use semantic data augmentation on the latent features to implicitly increase the diversity of the training sample. We conduct extensive experiments on five long-tailed image recognition datasets, and the results show that our proposed method significantly improves the baselines. |
| first_indexed | 2024-10-01T04:43:52Z |
| format | Journal Article |
| id | ntu-10356/170932 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T04:43:52Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | ntu-10356/1709322023-10-13T15:36:35Z LCReg: long-tailed image classification with latent categories based recognition Liu, Weide Wu, Zhonghua Wang, Yiming Ding, Henghui Liu, Fayao Lin, Jie Lin, Guosheng School of Computer Science and Engineering Engineering::Computer science and engineering Long-Tailed Image Classification In this work, we tackle the challenging problem of long-tailed image recognition. Previous long-tailed recognition approaches mainly focus on data augmentation or re-balancing strategies for the tail classes to give them more attention during model training. However, these methods are limited by the small number of training images for the tail classes, which results in poor feature representations. To address this issue, we propose the Latent Categories based long-tail Recognition (LCReg) method. Our hypothesis is that common latent features shared by head and tail classes can be used to improve feature representation. Specifically, we learn a set of class-agnostic latent features shared by both head and tail classes, and then use semantic data augmentation on the latent features to implicitly increase the diversity of the training sample. We conduct extensive experiments on five long-tailed image recognition datasets, and the results show that our proposed method significantly improves the baselines. Agency for Science, Technology and Research (A*STAR) Ministry of Education (MOE) Submitted/Accepted version This research is supported by the Ministry of Education, Singapore, under its Academic Research Fund Tier 1 (RG95/20). This research is also supported by the Agency for Science, Technology and Research (A*STAR) under its AME Programmatic Funds (Grant No. A20H6b0151). 2023-10-09T03:01:50Z 2023-10-09T03:01:50Z 2024 Journal Article Liu, W., Wu, Z., Wang, Y., Ding, H., Liu, F., Lin, J. & Lin, G. (2024). LCReg: long-tailed image classification with latent categories based recognition. Pattern Recognition, 145, 109971-. https://dx.doi.org/10.1016/j.patcog.2023.109971 0031-3203 https://hdl.handle.net/10356/170932 10.1016/j.patcog.2023.109971 2-s2.0-85171971903 145 109971 en RG95/20 A20H6b0151 Pattern Recognition © 2023 Elsevier Ltd. All rights reserved. This article may be downloaded for personal use only. Any other use requires prior permission of the copyright holder. The Version of Record is available online at http://doi.org/10.1016/j.patcog.2023.10. application/pdf |
| spellingShingle | Engineering::Computer science and engineering Long-Tailed Image Classification Liu, Weide Wu, Zhonghua Wang, Yiming Ding, Henghui Liu, Fayao Lin, Jie Lin, Guosheng LCReg: long-tailed image classification with latent categories based recognition |
| title | LCReg: long-tailed image classification with latent categories based recognition |
| title_full | LCReg: long-tailed image classification with latent categories based recognition |
| title_fullStr | LCReg: long-tailed image classification with latent categories based recognition |
| title_full_unstemmed | LCReg: long-tailed image classification with latent categories based recognition |
| title_short | LCReg: long-tailed image classification with latent categories based recognition |
| title_sort | lcreg long tailed image classification with latent categories based recognition |
| topic | Engineering::Computer science and engineering Long-Tailed Image Classification |
| url | https://hdl.handle.net/10356/170932 |
| work_keys_str_mv | AT liuweide lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition AT wuzhonghua lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition AT wangyiming lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition AT dinghenghui lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition AT liufayao lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition AT linjie lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition AT linguosheng lcreglongtailedimageclassificationwithlatentcategoriesbasedrecognition |