Research map / Detection and segmentation / Open-set and few-shot detection
DeFSS: Image-to-Mask Denoising Learning for Few-shot Segmentation
Zishu Qin, Junhao Xu, Weifeng Ge
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 image-to-mask denoising instead of image-to-mask decoding (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 68.8 -> 71.7 | Improved |
| use image-to-mask denoising instead of noise-to-mask denoising (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 64.4 -> 71.7 | Improved |
| add image corruption (naive) (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 67.8 -> 69.2 | Improved |
| make corruption adaptive with FG-BG-specific prior instead of naive (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 69.2 -> 71.7 | Improved |
| use FG-BG-specific prior generator instead of SCCAN prior (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 70.8 -> 71.7 | Improved |
| use corruption scale tau=400 instead of 800 (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 68.1 -> 71.7 | Improved |
| add condition ViT to the denoising network (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 65.2 -> 67.7 | Improved |
| add adaLN layers (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 67.7 -> 69.2 | Improved |
| add discriminative attribute learning with surrogates (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 69.2 -> 70.8 | Improved |
| use one-hot classification learning instead of surrogate attribute learning (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 70.8 -> 66.0 | Hurt |
| add cross-surrogate regularization SCE-Reg (ablation) | DeFSS · PASCAL-5i 1-shot, ResNet50 | mIoU: 70.8 -> 71.7 | Improved |
| use DeFSS instead of the best baseline in 1-shot vs 5-shot | DeFSS · PASCAL-5i and COCO-20i, ResNet50 | mIoU: +1.3 and 1.6 (1-shot), 1.0 and 1.1 (5-shot) | Improved |
Later papers that compared with this method
1 later accepted paper reports DeFSS as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Mitigating Pose–Scale Discrepancy Bias and Reforming Multi-Support Reasoning for Few-Shot Semantic Segmentation (ECCV 2026 · better on 8 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.89)
- Unleashing the Potential of the Diffusion Model in Few-shot Semantic Segmentation (NeurIPS 2024 · similarity 0.87)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 2024 · similarity 0.87)
- Masked Pre-training Enables Universal Zero-shot Denoiser (NeurIPS 2024 · similarity 0.87)
- Lightweight Frequency Masker for Cross-Domain Few-Shot Semantic Segmentation (NeurIPS 2024 · similarity 0.87)
- $\mathbf{D^{3}S^{2}}$: Diffusion-Guided Dataset Distillation for Semantic Segmentation (NeurIPS 2026 · similarity 0.86)
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.
