Research map / Detection and segmentation / Weakly supervised detection and segmentation
SSR: Semantic and Spatial Rectification for CLIP-based Weakly Supervised Segmentation
Xiuli Bi, Die Xiao, Junchao Fan, Bin Xiao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add CLIP multi-head attention affinity (ablation) | SSR CMPA CAM · PASCAL VOC train (CAM seeds) | mIoU (%): 63.3 -> 74.6 | Improved |
| add DINO multi-head attention to CLIP affinity (ablation) | SSR · PASCAL VOC train (CAM seeds) | mIoU (%): 74.6 -> 76.3 | Improved |
| add superpixel-guided correction (SGC) (ablation) | SSR · PASCAL VOC train (CAM seeds) | mIoU (%): 76.3 -> 78.7 | Improved |
| add direct feature fine-tuning loss (Lfeature) (ablation) | CLIP CAM · PASCAL VOC | mIoU (%): 58.6 -> 53.5 | Hurt |
| add intra-modal contrastive loss (ablation) | CLIP CAM · PASCAL VOC | mIoU (%): 58.6 -> 57.8 | Hurt |
| add cross-modal prototype contrastive loss (ablation) | CLIP CAM · PASCAL VOC | mIoU (%): 58.6 -> 63.3 | Improved |
| add CRF post-processing (ablation) | SSR · PASCAL VOC val | mIoU (%): 78.2 -> 79.5 | Improved |
| add CRF post-processing (ablation) | SSR · MS COCO val | mIoU (%): 49.2 -> 50.6 | Improved |
| use SSR instead of ExCEL | single-stage WSSS · PASCAL VOC val | mIoU (%): 78.4 -> 79.5 | Improved |
| use SSR instead of ExCEL | single-stage WSSS · MS COCO val | mIoU (%): 50.3 -> 50.6 | Improved |
| use SSR instead of VPL (multi-stage) | WSSS · MS COCO val | mIoU (%): 49.8 -> 50.6 | Improved |
Closest papers by meaning
- Exploring CLIP's Dense Knowledge for Weakly Supervised Semantic Segmentation (CVPR 2025 · similarity 0.89)
- Toward Modality Gap: Vision Prototype Learning for Weakly-supervised Semantic Segmentation with CLIP (AAAI 2025 · similarity 0.87)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.87)
- FFR: Frequency Feature Rectification for Weakly Supervised Semantic Segmentation (CVPR 2025 · similarity 0.86)
- Frozen CLIP: A Strong Backbone for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
