Research map / Detection and segmentation / Open-set and few-shot detection
Images as Noisy Labels: Unleashing the Potential of the Diffusion Model for Open-Vocabulary Semantic Segmentation
Fan Li, Xuanbin Wang, Xuan Wang, Zhaoxiang Zhang, Yuelei Xu
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 learnable proxy queries (single denoising step) to the baseline (ablation) | CLIP + Mask2Former + cost aggregation baseline · PASCAL Context PC-59 | mIoU: 60.6 -> 62.9 | Improved |
| add progressive learning (basic kernel x diffusion kernel over multiple denoising steps) (ablation) | DEDOS · PASCAL Context PC-59 | mIoU: 62.9 -> 64.2 | Improved |
| add CLIP-guided texture decoupling of diffusion features (ablation) | DEDOS · PASCAL Context PC-59 | mIoU: 64.2 -> 64.9 | Improved |
| add text interaction through proxy queries (ablation) | DEDOS · PASCAL Context PC-59 | mIoU: 64.9 -> 65.7 | Improved |
| swap Stable Diffusion version (1.4, 1.5, 2.1) (ablation) | DEDOS · PASCAL Context PC-59 | mIoU: 65.4 / 65.6 / 65.7 | No clear effect |
| increase denoising iterations beyond 3 (ablation) | DEDOS · PASCAL Context PC-59 | mIoU: 65.7 (3 and 5) -> 65.2 (7) | No clear effect |
| use larger diffusion timesteps [100, 200, 300] instead of [50, 100, 200] (ablation) | DEDOS · PASCAL Context PC-59 | mIoU: 65.7 -> 64.6 | Hurt |
| increase number or dimension of proxy queries beyond n=100 or c=256 | DEDOS · PASCAL Context PC-59 | mIoU | Hurt |
| DEDOS versus CAT-Seg on specialized domains (MESS medical and engineering) | DEDOS · MESS Engineering (ViT-B/16) | mIoU: DEDOS 22.6 vs SAN 23.6 | Mixed |
| use DEDOS instead of CAT-Seg for efficiency | DEDOS · Efficiency (ViT-L/14) | inference time (s): 0.37 -> 0.12 | Improved |
Closest papers by meaning
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · similarity 0.89)
- Unleashing the Potential of the Diffusion Model in Few-shot Semantic Segmentation (NeurIPS 2024 · similarity 0.89)
- Training-Free Open-Vocabulary Segmentation with Offline Diffusion-Augmented Prototype Generation (CVPR 2024 · similarity 0.89)
- Diffusion-Driven Two-Stage Active Learning for Low-Budget Semantic Segmentation (NeurIPS 2025 · similarity 0.89)
- Unsupervised Region-Based Image Editing of Denoising Diffusion Models (AAAI 2025 · similarity 0.88)
- EmerDiff: Emerging Pixel-level Semantic Knowledge in Diffusion Models (ICLR 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.
