Non-target feature filtering for weakly supervised semantic segmentation
Abstract Weakly supervised semantic segmentation (WSSS) utilizes weak labels to learn semantic segmentation models, significantly reducing reliance on pixel-level annotations. WSSS typically employs a multi-label classification network to extract image features for constructing localization maps. Th...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2024-12-01
|
| Series: | Complex & Intelligent Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s40747-024-01678-8 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832571200690192384 |
|---|---|
| author | Xuesheng Zhou Yan Li Guitao Cao Wenming Cao |
| author_facet | Xuesheng Zhou Yan Li Guitao Cao Wenming Cao |
| author_sort | Xuesheng Zhou |
| collection | DOAJ |
| description | Abstract Weakly supervised semantic segmentation (WSSS) utilizes weak labels to learn semantic segmentation models, significantly reducing reliance on pixel-level annotations. WSSS typically employs a multi-label classification network to extract image features for constructing localization maps. The quality of the localization map critically influences the performance of WSSS. However, non-target semantic noise within the features impedes the improvement of localization map quality. To address this issue, we propose a non-target feature filtering class activation mapping (NFF-CAM) for WSSS, which can reduce non-target semantic signals and generate higher-quality localization maps. Specifically, the class-constrained dual cosine clustering (CDCC) and channel identification (CI) modules are introduced in NFF-CAM. CDCC effectively addresses the issue of unsuitability in the clustering group relationships of the original features under specified class conditions. CI can efficiently identify channel features containing non-target semantic information. We conduct extensive evaluations of NFF-CAM on popular datasets, including PASCAL VOC 2012 and MS COCO 2014. Experimental results show that NFF-CAM can effectively improve the segmentation performance of off-the-shelf methods. |
| format | Article |
| id | doaj-art-d3e70e0557ee4ec2aa5b1effe6087ab9 |
| institution | Kabale University |
| issn | 2199-4536 2198-6053 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Springer |
| record_format | Article |
| series | Complex & Intelligent Systems |
| spelling | doaj-art-d3e70e0557ee4ec2aa5b1effe6087ab92025-02-02T12:49:17ZengSpringerComplex & Intelligent Systems2199-45362198-60532024-12-0111111510.1007/s40747-024-01678-8Non-target feature filtering for weakly supervised semantic segmentationXuesheng Zhou0Yan Li1Guitao Cao2Wenming Cao3MoE Engineering Research Center of SW/HW Co-design Technology and Application, East China Normal UniversityDepartment of Computer Science, Shanghai Normal UniversityMoE Engineering Research Center of SW/HW Co-design Technology and Application, East China Normal UniversityCollege of Information Engineering, Shenzhen UniversityAbstract Weakly supervised semantic segmentation (WSSS) utilizes weak labels to learn semantic segmentation models, significantly reducing reliance on pixel-level annotations. WSSS typically employs a multi-label classification network to extract image features for constructing localization maps. The quality of the localization map critically influences the performance of WSSS. However, non-target semantic noise within the features impedes the improvement of localization map quality. To address this issue, we propose a non-target feature filtering class activation mapping (NFF-CAM) for WSSS, which can reduce non-target semantic signals and generate higher-quality localization maps. Specifically, the class-constrained dual cosine clustering (CDCC) and channel identification (CI) modules are introduced in NFF-CAM. CDCC effectively addresses the issue of unsuitability in the clustering group relationships of the original features under specified class conditions. CI can efficiently identify channel features containing non-target semantic information. We conduct extensive evaluations of NFF-CAM on popular datasets, including PASCAL VOC 2012 and MS COCO 2014. Experimental results show that NFF-CAM can effectively improve the segmentation performance of off-the-shelf methods.https://doi.org/10.1007/s40747-024-01678-8Semantic segmentationWeak supervisionLocalization mapClass activation mappingClassification |
| spellingShingle | Xuesheng Zhou Yan Li Guitao Cao Wenming Cao Non-target feature filtering for weakly supervised semantic segmentation Complex & Intelligent Systems Semantic segmentation Weak supervision Localization map Class activation mapping Classification |
| title | Non-target feature filtering for weakly supervised semantic segmentation |
| title_full | Non-target feature filtering for weakly supervised semantic segmentation |
| title_fullStr | Non-target feature filtering for weakly supervised semantic segmentation |
| title_full_unstemmed | Non-target feature filtering for weakly supervised semantic segmentation |
| title_short | Non-target feature filtering for weakly supervised semantic segmentation |
| title_sort | non target feature filtering for weakly supervised semantic segmentation |
| topic | Semantic segmentation Weak supervision Localization map Class activation mapping Classification |
| url | https://doi.org/10.1007/s40747-024-01678-8 |
| work_keys_str_mv | AT xueshengzhou nontargetfeaturefilteringforweaklysupervisedsemanticsegmentation AT yanli nontargetfeaturefilteringforweaklysupervisedsemanticsegmentation AT guitaocao nontargetfeaturefilteringforweaklysupervisedsemanticsegmentation AT wenmingcao nontargetfeaturefilteringforweaklysupervisedsemanticsegmentation |