A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation
The key to semi-supervised semantic segmentation is to assign the appropriate pseudo-label to the pixels of unlabeled images. Recently, various approaches to consistency-based training and the filtering of reliable pseudo-labels have shown remarkable results. Nonetheless, there are still issues to b...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-12-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/13/24/13255 |
| _version_ | 1827575785391652864 |
|---|---|
| author | Hyunmin Gwak Yongho Jeong Chanyeong Kim Yonghak Lee Seongmin Yang Sunghwan Kim |
| author_facet | Hyunmin Gwak Yongho Jeong Chanyeong Kim Yonghak Lee Seongmin Yang Sunghwan Kim |
| author_sort | Hyunmin Gwak |
| collection | DOAJ |
| description | The key to semi-supervised semantic segmentation is to assign the appropriate pseudo-label to the pixels of unlabeled images. Recently, various approaches to consistency-based training and the filtering of reliable pseudo-labels have shown remarkable results. Nonetheless, there are still issues to be addressed. We find that recent approaches have specific problems in common. In pseudo-labels for training unlabeled images, we confirm that false foreground class pseudo-labels are mostly caused by background class confusion, not confusion between different foreground classes. To solve this problem, we propose a foreground and background discrimination model for semi-supervised semantic segmentation. Our proposed model is trained using a novel approach called multi-view integrated ensemble (MVIE) via output perturbation. Experimental results in various partition protocols show that our approach outperforms the existing state of the art (SOTA) in binary prediction on unlabeled data, and the segmentation model trained with the help of our model outperforms existing models. |
| first_indexed | 2024-03-08T21:01:55Z |
| format | Article |
| id | doaj.art-bd619bd955ca43c581be2bd1529b6dce |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-08T21:01:55Z |
| publishDate | 2023-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-bd619bd955ca43c581be2bd1529b6dce2023-12-22T13:52:02ZengMDPI AGApplied Sciences2076-34172023-12-0113241325510.3390/app132413255A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic SegmentationHyunmin Gwak0Yongho Jeong1Chanyeong Kim2Yonghak Lee3Seongmin Yang4Sunghwan Kim5Department of Applied Statistics, Konkuk University, Seoul 05029, Republic of KoreaAI Analytics Team, Mustree, Seoul 05029, Republic of KoreaDepartment of Applied Statistics, Konkuk University, Seoul 05029, Republic of KoreaDepartment of Applied Statistics, Konkuk University, Seoul 05029, Republic of KoreaDepartment of Applied Statistics, Konkuk University, Seoul 05029, Republic of KoreaDepartment of Applied Statistics, Konkuk University, Seoul 05029, Republic of KoreaThe key to semi-supervised semantic segmentation is to assign the appropriate pseudo-label to the pixels of unlabeled images. Recently, various approaches to consistency-based training and the filtering of reliable pseudo-labels have shown remarkable results. Nonetheless, there are still issues to be addressed. We find that recent approaches have specific problems in common. In pseudo-labels for training unlabeled images, we confirm that false foreground class pseudo-labels are mostly caused by background class confusion, not confusion between different foreground classes. To solve this problem, we propose a foreground and background discrimination model for semi-supervised semantic segmentation. Our proposed model is trained using a novel approach called multi-view integrated ensemble (MVIE) via output perturbation. Experimental results in various partition protocols show that our approach outperforms the existing state of the art (SOTA) in binary prediction on unlabeled data, and the segmentation model trained with the help of our model outperforms existing models.https://www.mdpi.com/2076-3417/13/24/13255deep learningimage semantic segmentationsemi-supervised semantic segmentationreliable pseudo-labels |
| spellingShingle | Hyunmin Gwak Yongho Jeong Chanyeong Kim Yonghak Lee Seongmin Yang Sunghwan Kim A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation Applied Sciences deep learning image semantic segmentation semi-supervised semantic segmentation reliable pseudo-labels |
| title | A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation |
| title_full | A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation |
| title_fullStr | A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation |
| title_full_unstemmed | A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation |
| title_short | A Multi-View Integrated Ensemble for the Background Discrimination of Semi-Supervised Semantic Segmentation |
| title_sort | multi view integrated ensemble for the background discrimination of semi supervised semantic segmentation |
| topic | deep learning image semantic segmentation semi-supervised semantic segmentation reliable pseudo-labels |
| url | https://www.mdpi.com/2076-3417/13/24/13255 |
| work_keys_str_mv | AT hyunmingwak amultiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT yonghojeong amultiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT chanyeongkim amultiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT yonghaklee amultiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT seongminyang amultiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT sunghwankim amultiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT hyunmingwak multiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT yonghojeong multiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT chanyeongkim multiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT yonghaklee multiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT seongminyang multiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation AT sunghwankim multiviewintegratedensembleforthebackgrounddiscriminationofsemisupervisedsemanticsegmentation |