Semi-supervised and long-tailed object detection with Cascadematch
This paper focuses on long-tailed object detection in the semi-supervised learning setting, which poses realistic challenges, but has rarely been studied in the literature. We propose a novel pseudo-labeling-based detector called CascadeMatch. Our detector features a cascade network architecture, wh...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/165791 |
| _version_ | 1826121385061122048 |
|---|---|
| author | Zang, Yuhang Zhou, Kaiyang Huang, Chen Loy, Chen Change |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Zang, Yuhang Zhou, Kaiyang Huang, Chen Loy, Chen Change |
| author_sort | Zang, Yuhang |
| collection | NTU |
| description | This paper focuses on long-tailed object detection in the semi-supervised learning setting, which poses realistic challenges, but has rarely been studied in the literature. We propose a novel pseudo-labeling-based detector called CascadeMatch. Our detector features a cascade network architecture, which has multi-stage detection heads with progressive confidence thresholds. To avoid manually tuning the thresholds, we design a new adaptive pseudo-label mining mechanism to automatically identify suitable values from data. To mitigate confirmation bias, where a model is negatively reinforced by incorrect pseudo-labels produced by itself, each detection head is trained by the ensemble pseudo-labels of all detection heads. Experiments on two long-tailed datasets, i.e., LVIS and COCO-LT, demonstrate that CascadeMatch surpasses existing state-of-the-art semi-supervised approaches—across a wide range of detection architectures—in handling long-tailed object detection. For instance, CascadeMatch outperforms Unbiased Teacher by 1.9 AP Fix on LVIS when using a ResNet50-based Cascade R-CNN structure, and by 1.7 AP Fix when using Sparse R-CNN with a Transformer encoder. We also show that CascadeMatch can even handle the challenging sparsely annotated object detection problem. Code: https://github.com/yuhangzang/CascadeMatch. |
| first_indexed | 2024-10-01T05:31:47Z |
| format | Journal Article |
| id | ntu-10356/165791 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T05:31:47Z |
| publishDate | 2023 |
| record_format | dspace |
| spelling | ntu-10356/1657912023-04-14T15:36:08Z Semi-supervised and long-tailed object detection with Cascadematch Zang, Yuhang Zhou, Kaiyang Huang, Chen Loy, Chen Change School of Computer Science and Engineering S-Lab, NTU Engineering::Computer science and engineering Object Detection Long-Tailed Learning This paper focuses on long-tailed object detection in the semi-supervised learning setting, which poses realistic challenges, but has rarely been studied in the literature. We propose a novel pseudo-labeling-based detector called CascadeMatch. Our detector features a cascade network architecture, which has multi-stage detection heads with progressive confidence thresholds. To avoid manually tuning the thresholds, we design a new adaptive pseudo-label mining mechanism to automatically identify suitable values from data. To mitigate confirmation bias, where a model is negatively reinforced by incorrect pseudo-labels produced by itself, each detection head is trained by the ensemble pseudo-labels of all detection heads. Experiments on two long-tailed datasets, i.e., LVIS and COCO-LT, demonstrate that CascadeMatch surpasses existing state-of-the-art semi-supervised approaches—across a wide range of detection architectures—in handling long-tailed object detection. For instance, CascadeMatch outperforms Unbiased Teacher by 1.9 AP Fix on LVIS when using a ResNet50-based Cascade R-CNN structure, and by 1.7 AP Fix when using Sparse R-CNN with a Transformer encoder. We also show that CascadeMatch can even handle the challenging sparsely annotated object detection problem. Code: https://github.com/yuhangzang/CascadeMatch. Ministry of Education (MOE) Nanyang Technological University Submitted/Accepted version This study is supported under the RIE2020 Industry Alignment Fund Industry Collaboration Projects (IAF-ICP) Funding Initiative, as well as cash and in-kind contribution from the industry partner(s). It is also partly supported by the NTU NAP grant and Singapore MOE AcRF Tier 2 (MOE-T2EP20120-0001). 2023-04-10T08:33:03Z 2023-04-10T08:33:03Z 2023 Journal Article Zang, Y., Zhou, K., Huang, C. & Loy, C. C. (2023). Semi-supervised and long-tailed object detection with Cascadematch. International Journal of Computer Vision, 131(4), 987-1001. https://dx.doi.org/10.1007/s11263-022-01738-x 0920-5691 https://hdl.handle.net/10356/165791 10.1007/s11263-022-01738-x 2-s2.0-85145831630 4 131 987 1001 en MOE-T2EP20120-0001 NTU NAP International Journal of Computer Vision © 2023 The Author(s), under exclusive licence to Springer Science + Business Media, LLC, part of Springer Nature. All rights reserved. This version of the article has been accepted for publication, after peer review and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: http://dx.doi.org/10.1007/s11263-022-01738-x application/pdf |
| spellingShingle | Engineering::Computer science and engineering Object Detection Long-Tailed Learning Zang, Yuhang Zhou, Kaiyang Huang, Chen Loy, Chen Change Semi-supervised and long-tailed object detection with Cascadematch |
| title | Semi-supervised and long-tailed object detection with Cascadematch |
| title_full | Semi-supervised and long-tailed object detection with Cascadematch |
| title_fullStr | Semi-supervised and long-tailed object detection with Cascadematch |
| title_full_unstemmed | Semi-supervised and long-tailed object detection with Cascadematch |
| title_short | Semi-supervised and long-tailed object detection with Cascadematch |
| title_sort | semi supervised and long tailed object detection with cascadematch |
| topic | Engineering::Computer science and engineering Object Detection Long-Tailed Learning |
| url | https://hdl.handle.net/10356/165791 |
| work_keys_str_mv | AT zangyuhang semisupervisedandlongtailedobjectdetectionwithcascadematch AT zhoukaiyang semisupervisedandlongtailedobjectdetectionwithcascadematch AT huangchen semisupervisedandlongtailedobjectdetectionwithcascadematch AT loychenchange semisupervisedandlongtailedobjectdetectionwithcascadematch |