A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network
Due to the high cost and time-consumption of artificial semantic tags, domain-based adaptive semantics segmentation is very necessary. For scenes with large gaps or pixels, it is easy to limit model training and reduce the accuracy of semantic segmentation. In this paper, a domain adaptive semantic...
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Journal of Shanghai Jiao Tong University
2021-09-01
|
| Series: | Shanghai Jiaotong Daxue xuebao |
| Subjects: | |
| Online Access: | http://xuebao.sjtu.edu.cn/article/2021/1006-2467/1006-2467-55-9-1158.shtml |
| _version_ | 1818693298326339584 |
|---|---|
| author | ZHANG Junning, SU Qunxing, WANG Cheng, XU Chao, LI Yining |
| author_facet | ZHANG Junning, SU Qunxing, WANG Cheng, XU Chao, LI Yining |
| author_sort | ZHANG Junning, SU Qunxing, WANG Cheng, XU Chao, LI Yining |
| collection | DOAJ |
| description | Due to the high cost and time-consumption of artificial semantic tags, domain-based adaptive semantics segmentation is very necessary. For scenes with large gaps or pixels, it is easy to limit model training and reduce the accuracy of semantic segmentation. In this paper, a domain adaptive semantic segmentation network (DA-SSN) using the improved transformation network is proposed by eliminating the interference of large gap pictures and pixels through staged training and interpretable masks. First, in view of the problem of large domain gaps from some source graphs to target graphs and the difficulty in network model training, the training loss threshold is used to divide the source graph dataset with large gaps, and a phased transformation network training strategy is proposed. Based on the ensurance of the semantic alignment of small gap source images, the transformation quality of large gap source images is improved. In addition, in order to further reduce the gap between some pixels in the source image and the target image area, an interpretable mask is proposed. By predicting the gap between each pixel in the source image domain and the target image domain, the confidence is reduced, and the training loss of the corresponding pixel is ignored to eliminate the influence of large gap pixels on the semantic alignment of other pixels, so that model training only focuses on the domain gap of high-confidence pixels. The results show that the proposed algorithm has a higher segmentation accuracy than the original domain adaptive semantic segmentation network. Compared with the results of other popular algorithms, the proposed method obtains a higher quality semantic alignment, which shows the advantages of the proposed method with high accuracy. |
| first_indexed | 2024-12-17T13:11:28Z |
| format | Article |
| id | doaj.art-67a0f41611ec4f3dbd2a6d8297370791 |
| institution | Directory Open Access Journal |
| issn | 1006-2467 |
| language | zho |
| last_indexed | 2024-12-17T13:11:28Z |
| publishDate | 2021-09-01 |
| publisher | Editorial Office of Journal of Shanghai Jiao Tong University |
| record_format | Article |
| series | Shanghai Jiaotong Daxue xuebao |
| spelling | doaj.art-67a0f41611ec4f3dbd2a6d82973707912022-12-21T21:47:06ZzhoEditorial Office of Journal of Shanghai Jiao Tong UniversityShanghai Jiaotong Daxue xuebao1006-24672021-09-015591158116810.16183/j.cnki.jsjtu.2019.307A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation NetworkZHANG Junning, SU Qunxing, WANG Cheng, XU Chao, LI Yining01. Army Engineering University, Shijiazhuang 050003, China;2. Army Command College, Nanjing 210000, China;3. Joint Operations Academy, National Defense University, Shijiazhuang 050003, China;4. 32181 Troops, Xi’an 710032, China;5. Research Institute of Chemical Defense, Academy of Military Sciences, Beijing 102205, ChinaDue to the high cost and time-consumption of artificial semantic tags, domain-based adaptive semantics segmentation is very necessary. For scenes with large gaps or pixels, it is easy to limit model training and reduce the accuracy of semantic segmentation. In this paper, a domain adaptive semantic segmentation network (DA-SSN) using the improved transformation network is proposed by eliminating the interference of large gap pictures and pixels through staged training and interpretable masks. First, in view of the problem of large domain gaps from some source graphs to target graphs and the difficulty in network model training, the training loss threshold is used to divide the source graph dataset with large gaps, and a phased transformation network training strategy is proposed. Based on the ensurance of the semantic alignment of small gap source images, the transformation quality of large gap source images is improved. In addition, in order to further reduce the gap between some pixels in the source image and the target image area, an interpretable mask is proposed. By predicting the gap between each pixel in the source image domain and the target image domain, the confidence is reduced, and the training loss of the corresponding pixel is ignored to eliminate the influence of large gap pixels on the semantic alignment of other pixels, so that model training only focuses on the domain gap of high-confidence pixels. The results show that the proposed algorithm has a higher segmentation accuracy than the original domain adaptive semantic segmentation network. Compared with the results of other popular algorithms, the proposed method obtains a higher quality semantic alignment, which shows the advantages of the proposed method with high accuracy.http://xuebao.sjtu.edu.cn/article/2021/1006-2467/1006-2467-55-9-1158.shtmlcomputer visiondomain adaptive semantics segmentationdomain gapsemantic information integrationinterpretable maskstaged training |
| spellingShingle | ZHANG Junning, SU Qunxing, WANG Cheng, XU Chao, LI Yining A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network Shanghai Jiaotong Daxue xuebao computer vision domain adaptive semantics segmentation domain gap semantic information integration interpretable mask staged training |
| title | A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network |
| title_full | A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network |
| title_fullStr | A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network |
| title_full_unstemmed | A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network |
| title_short | A Domain Adaptive Semantic Segmentation Network Based on Improved Transformation Network |
| title_sort | domain adaptive semantic segmentation network based on improved transformation network |
| topic | computer vision domain adaptive semantics segmentation domain gap semantic information integration interpretable mask staged training |
| url | http://xuebao.sjtu.edu.cn/article/2021/1006-2467/1006-2467-55-9-1158.shtml |
| work_keys_str_mv | AT zhangjunningsuqunxingwangchengxuchaoliyining adomainadaptivesemanticsegmentationnetworkbasedonimprovedtransformationnetwork AT zhangjunningsuqunxingwangchengxuchaoliyining domainadaptivesemanticsegmentationnetworkbasedonimprovedtransformationnetwork |