Research map / Detection and segmentation / CNN features
DifNet: Semantic Segmentation by Diffusion Networks
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 |
|---|---|---|---|
| replace Sim-Deeplab-18 with DifNet-18 | ResNet-18 semantic segmentation on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 66.33% -> 70.17% | Helped |
| replace Sim-Deeplab-34 with DifNet-34 | ResNet-34 semantic segmentation on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 69.76% -> 71.84% | Helped |
| replace Sim-Deeplab-50 with DifNet-50 | ResNet-50 semantic segmentation on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 70.78% -> 72.57% | Helped |
| replace Sim-Deeplab-101 with DifNet-101 | ResNet-101 semantic segmentation on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 71.83% -> 73.22% | Helped |
| replace Sim-Deeplab-101 with DifNet-50 | semantic segmentation on Pascal VOC 2012 test · Pascal VOC 2012 test | mIOU: 72.54% -> 72.55% | Helped |
| replace Sim-Deeplab-101 with DifNet-101 | ResNet-101 semantic segmentation on Pascal VOC 2012 test · Pascal VOC 2012 test | mIOU: 72.54% -> 73.21% | Helped |
| add diffusion to the ResNet-101 Deeplab configuration with MSC, COCO pretraining, ASPP, and CRF | Pascal Context segmentation · Pascal Context val | mIOU: 45.7% -> 46.0% | Helped |
| remove ASPP (ablation) | DifNet-50 on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 72.52% -> 72.57% | No clear effect |
| remove the 2nd and 4th random walks (ablation) | DifNet-50 on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 1 percent drop | Hurt |
| add CRF post-processing | Sim-Deeplab-101 on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 71.83% -> 72.26% | Helped |
| replace Sim-Deeplab-101-CRF with DifNet-50 | semantic segmentation on Pascal VOC 2012 val · Pascal VOC 2012 val | mIOU: 72.26% -> 72.57% | Helped |
Closest papers by meaning
- DiffCut: Catalyzing Zero-Shot Semantic Segmentation with Diffusion Features and Recursive Normalized Cut (NeurIPS 2024 · similarity 0.89)
- Unleashing the Potential of the Diffusion Model in Few-shot Semantic Segmentation (NeurIPS 2024 · similarity 0.88)
- $\mathbf{D^{3}S^{2}}$: Diffusion-Guided Dataset Distillation for Semantic Segmentation (NeurIPS 2026 · similarity 0.88)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · similarity 0.87)
- EmerDiff: Emerging Pixel-level Semantic Knowledge in Diffusion Models (ICLR 2024 · similarity 0.87)
- SeeDiff: Off-the-Shelf Seeded Mask Generation from Diffusion Models (AAAI 2025 · 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.