Research map / Detection and segmentation / Weakly supervised detection and segmentation
Self-Supervised Difference Detection for Weakly-Supervised Semantic Segmentation
Wataru Shimoda, Keiji Yanai
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 SSDD dynamic region refinement on top of PSA | PSA ResNet-38 · PASCAL VOC 2012 val (image-level labels only) | mIoU: 61.7 -> 64.9 | Improved |
| add SSDD refinement on top of PSA, test set | PSA · PASCAL VOC 2012 test (image-level labels only) | mIoU: 63.7 -> 65.5 | Improved |
| add SSDD to animal classes (cat, cow, horse, sheep) | PSA · PASCAL VOC 2012 val (image-level labels only) | per-class IoU: 75.1 -> 87.0 (cat) | Improved |
| add SSDD to potted plant and person classes | PSA · PASCAL VOC 2012 val (image-level labels only) | per-class IoU: 42.5 -> 29.8 (plant); 73.7 -> 69.6 (person) | Hurt |
| set bias b_dd to 0.4 versus 0.0 (ablation) | SSDD · PASCAL VOC 2012 val, hyperparameter sweep | mIoU: 62.2 -> 64.9 | Improved |
| set bias b_dd to 0.5 versus 0.4 (ablation) | SSDD · PASCAL VOC 2012 val, hyperparameter sweep | mIoU: 64.9 -> 62.7 | Hurt |
| set class bias b_class to 1.0 versus 0.0 (ablation) | SSDD · PASCAL VOC 2012 val, hyperparameter sweep | mIoU: 64.3 -> 64.9 | Improved |
| set class bias b_class to 0.5 versus 1.0 (ablation) | SSDD · PASCAL VOC 2012 val, hyperparameter sweep | mIoU: 64.9 -> 63.0 | Hurt |
| set mixing ratio alpha to 0.5 versus using only one mask (alpha 1.0) (ablation) | SSDD · PASCAL VOC 2012 val, hyperparameter sweep | mIoU: 63.1 -> 64.9 | Improved |
| set mixing ratio alpha to 0.5 versus alpha 0.0 (ablation) | SSDD · PASCAL VOC 2012 val, hyperparameter sweep | mIoU: 63.2 -> 64.9 | Improved |
| use SSDD without any extra supervision versus methods using saliency masks | SSDD · PASCAL VOC 2012 val (image-level labels only) | mIoU: 61.4 -> 64.9 | Improved |
Closest papers by meaning
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.90)
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.90)
- Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning (ICLR 2021 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.88)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.88)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.87)
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.
