Research map / Detection and segmentation / Weakly supervised detection and segmentation
Simple Does It: Weakly Supervised Instance and Semantic Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 ignore regions using a 20% inner-box foreground label (ablation) | Box recursive training · VOC12 val | mIoU: 61.2 -> 62.7 | Helped |
| Combine MCG proposals with GrabCut+ and label only their agreement as foreground (ablation) | Weakly supervised semantic segmentation · VOC12 val | mIoU: 63.4 -> 65.7 | Helped |
| Use the MCG ∩ GrabCut+ input labels rather than Boxᶦ (ablation) | Weakly supervised semantic segmentation · VOC12 test | mIoU: 63.5 -> 67.5 | Helped |
| Add Pascal ground-truth masks to box-supervised VOC12 training (ablation) | MCG ∩ GrabCut+ semi-supervised training · VOC12 val | mIoU: 65.7 versus 65.8 | No clear effect |
| Add COCO bounding-box training data (ablation) | MCG ∩ GrabCut+ with Pascal ground-truth supervision · VOC12 val | mIoU: 69.1 -> 71.6 | Helped |
| Use weakly supervised HED boundaries instead of BSDS500-trained HED (ablation) | MCG ∩ GrabCut+ · VOC12 val | mIoU: 65.7 -> 64.8 | Hurt |
| Replace GrabCut RGB pairwise terms with HED boundary probabilities (ablation) | GrabCut box-guided segmentation · Pascal VOC12 val box-guided segments | mIoU: +~2 points | Helped |
| Use naive recursive training without the proposed post-processing (ablation) | Box recursive training · VOC12 val | mIoU | Hurt |
| Apply box enforcing and outlier reset between recursive rounds (ablation) | Naive box recursive training · VOC12 val | mIoU: 52.5 -> 55.6 | Helped |
| Add DenseCRF filtering between recursive training rounds (ablation) | Box recursive training with box enforcing and outlier reset · VOC12 val | mIoU | Helped |
| Train DeepLab_BOX instead of DeepMask | Weak instance segmentation · VOC12 val | mAPᵣ@0.5: 39.4 -> 44.8 | Helped |
| Use GrabCut+ instead of GrabCut for box-driven instance masks (ablation) | Training-free instance segmentation baseline · VOC12 val | mAPᵣ@0.5: 38.5 -> 41.1 | Helped |
Closest papers by meaning
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025 · similarity 0.91)
- Weakly Supervised Instance Segmentation using the Bounding Box Tightness Prior (NeurIPS 2019 · similarity 0.91)
- Semi-Supervised Semantic Image Segmentation With Self-Correcting Networks (CVPR 2020 · similarity 0.91)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.89)
- Sketchy Bounding-box Supervision for 3D Instance Segmentation (CVPR 2025 · 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.