Research map / Detection and segmentation / Weakly supervised detection and segmentation
Weakly-Supervised Semantic Segmentation with Visual Words Learning and Hybrid Pooling
Lixiang Ru, Bo Du, Yibing Zhan, Chen Wu
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 hybrid pooling (ablation) | ResNet101 baseline · PASCAL VOC train | CAM mIoU: 49.5 -> 54.0 | Improved |
| add visual words learning (memory-bank) on top of HP (ablation) | HP ResNet101 · PASCAL VOC train | CAM mIoU: 54.0 -> 56.9 | Improved |
| add visual words learning (learning-based, with DeCov) on top of HP (ablation) | HP ResNet101 · PASCAL VOC train | CAM mIoU: 54.0 -> 57.3 | Improved |
| add DeCov regularization to the learning-based codebook (ablation) | VWL-L · PASCAL VOC train | CAM mIoU: 55.1 -> 57.3 | Improved |
| use random sample codebook initialization instead of random (ablation) | VWL-L · PASCAL VOC train | CAM mIoU: 57.3 -> 55.8 | Hurt |
| use random sample codebook initialization instead of random (ablation) | VWL-M · PASCAL VOC train | CAM mIoU: 56.9 -> 56.6 | No clear effect |
| add GAP into hybrid pooling (ablation) | HP · PASCAL VOC train | CAM mIoU: about 16% mIoU improvement | Improved |
| use multi-scale split sizes {1,2,4} instead of {1} (ablation) | HP · PASCAL VOC train | CAM mIoU: 56.2 -> 57.3 | Improved |
| add split size 8 to {1,2,4} (ablation) | HP · PASCAL VOC train | CAM mIoU: 57.3 -> 57.0 | No clear effect |
| replace GAP with hybrid pooling (ablation) | ResNet101 baseline · PASCAL VOC train | CAM mIoU: 49.5 -> 54.0 | Improved |
| use GAP rather than HP to predict visual words (ablation) | VWL · PASCAL VOC train | CAM mIoU: 55.5 -> 57.3 | Improved |
| memory-bank vs learning-based strategy for visual word quality (ablation) | VWL · PASCAL VOC train classification accuracy | classification accuracy (%): 81.3 -> 84.8 | Improved |
Closest papers by meaning
- Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation (CVPR 2021 · similarity 0.91)
- Complementary Patch for Weakly Supervised Semantic Segmentation (ICCV 2021 · similarity 0.90)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.90)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.89)
- Separate and Conquer: Decoupling Co-occurrence via Decomposition and Representation for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.89)
- From SAM to CAMs: Exploring Segment Anything Model for Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.89)
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.
