Research map / Detection and segmentation / Open-set and few-shot detection
Weak-shot Semantic Segmentation via Dual Similarity Transfer
Read the paperCheck 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 pixel-pixel similarity transfer to proposal-only model (ablation) | SimFormer · COCO-Stuff-10K Split1 | novel-class mIoU: 23.5 -> 27.4 | Helped |
| Add complementary loss to proposal-only model (ablation) | SimFormer · COCO-Stuff-10K Split1 | novel-class mIoU: 23.5 -> 27.0 | Helped |
| Add complementary loss to Pr+Pi (full model) (ablation) | SimFormer · COCO-Stuff-10K Split1 | novel-class mIoU: 27.4 -> 31.1 | Helped |
| Add pixel-pixel transfer to Pr+Co (full model) (ablation) | SimFormer · COCO-Stuff-10K Split4 | novel-class mIoU: 25.6 -> 29.5 | Helped |
| Append re-training stage with mixed labels | SimFormer · COCO-Stuff-10K Split1 | novel-class mIoU: 31.1 -> 33.5 | Helped |
| Append re-training stage with mixed labels | SimFormer · ADE20K Split4 | novel-class mIoU: 28.5 -> 31.8 | Helped |
| Replace RETAB with SimFormer (both with re-training) | RETAB* · COCO-Stuff-10K Split4 | novel-class mIoU: 25.6 -> 31.7 | Helped |
| Construct pixel pairs within image (Self-Pair) instead of none (ablation) | Basic Model · COCO-Stuff-10K Split1 | novel-class mIoU: 27.0 -> 27.8 | Helped |
| Construct pixel pairs across images (Cross-Pair) instead of within image (ablation) | Self-Pair · COCO-Stuff-10K Split4 | novel-class mIoU: 26.6 -> 29.5 | Helped |
| Increase sampled pixel number J from 50 to 150 (ablation) | SimFormer · COCO-Stuff-10K Split2 | novel-class mIoU: 25.6 -> 26.3 | Helped |
| Increase novel-class proportion from 25% to 50% (Split 4 to Split 9) | SimFormer · COCO-Stuff-10K | novel-class mIoU: 29.5 -> 23.2 | Hurt |
| Increase gamma (ignore-class prior) (ablation) | SimFormer · COCO-Stuff-10K | novel-class mIoU | Hurt |
Closest papers by meaning
- Reliability Does Matter: An End-to-End Weakly Supervised Semantic Segmentation Approach (AAAI 2020 · similarity 0.91)
- Box-Driven Class-Wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation (CVPR 2019 · similarity 0.91)
- Bootstrapping the Performance of Webly Supervised Semantic Segmentation (CVPR 2018 · similarity 0.90)
- DuPL: Dual Student with Trustworthy Progressive Learning for Robust Weakly Supervised Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Progressive Feature Self-Reinforcement for Weakly Supervised Semantic Segmentation (AAAI 2024 · similarity 0.88)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 2024 · similarity 0.88)
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.