Research map / Detection and segmentation / Weakly supervised detection and segmentation
Learning to Detect Semantic Boundaries with Image-Level Class Labels
Namyup Kim, Sehyun Hwang, Suha Kwak
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 |
|---|---|---|---|
| use WSBDN architecture instead of Class-aware IRNet | pseudo label generation · SBD train soft boundary | MF (%): 59.7 -> 62.4 | Improved |
| use WSBDN architecture instead of WS-CASENet | pseudo label generation · SBD train hard boundary | MF (%): 56.0 -> 59.8 | Improved |
| multiply class-agnostic branch with class-aware branch (ablation) | WSBDN · SBD train class-aware | mAP (%): 46.6 -> 58.0 | Improved |
| multiply class-aware branch into class-agnostic branch (ablation) | WSBDN · SBD train class-agnostic | MF (%): 58.1 -> 68.2 | Improved |
| add multi-scale and flip inference (ablation) | WSBDN · SBD train hard boundary | MF (%): 56.5 -> 57.7 | Improved |
| add NMS without multi-scale/flip (ablation) | WSBDN · SBD train hard boundary | MF (%): 56.5 -> 55.5 | Hurt |
| add NMS on top of multi-scale/flip (ablation) | WSBDN · SBD train hard boundary | MF (%): 57.7 -> 59.8 | Improved |
| degrade CAM quality (IoU drops by 10%) (ablation) | WSBDN · SBD train pseudo labels | MF: drops by 4.5% | Hurt |
| use ground-truth segmentation masks instead of CAM (ablation) | WSBDN · SBD train pseudo labels | MF (%): up to 74.4% | Improved |
| vary line-segment gamma at 10 or above (ablation) | WSBDN · SBD train pseudo labels | MF: varying less than 3% | No clear effect |
| replace CASENet with DFF trained on pseudo labels | Ours · SBD test | mean MF (%): 59.7 -> 60.7 | Improved |
| use ResNet50 instead of VGG16 backbone | Ours+CASENet · SBD test | mean MF (%): 50.2 -> 59.8 | Improved |
Closest papers by meaning
- Devil Is in the Edges: Learning Semantic Boundaries From Noisy Annotations (CVPR 2019 · similarity 0.90)
- Weakly Supervised Learning of Instance Segmentation With Inter-Pixel Relations (CVPR 2019 · similarity 0.89)
- Weakly Supervised Semantic Segmentation with Boundary Exploration (ECCV 2020 · similarity 0.88)
- Large Scale Visual Recognition through Adaptation using Joint Representation and Multiple Instance Learning (JMLR 2016 · similarity 0.88)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.87)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 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.
