Research map / Image generation / Fast diffusion sampling
DGInStyle: Domain-Generalizable Semantic Segmentation with Image Diffusion Models and Stylized Semantic Control
Yuru Jia, Lukas Hoyer, Shengyu Huang, Tianfu Wang, Luc Van Gool, Konrad Schindler, Anton Obukhov
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 DGInStyle generated data to DG methods trained on GTA | IBN-Net, RobustNet, HRDA, Color-Aug, DAFormer · GTA to 5 real datasets | mIoU (Avg5): +2.5 to +7.2 mIoU | Improved |
| add DGInStyle data when source is Cityscapes | Color-Aug, DAFormer, HRDA · Cityscapes to BDD, MV, ACDC, DZ | mIoU (average): +2.1, +1.5, +0.7 | Improved |
| add Multi-Resolution Latent Fusion (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 43.31 -> 44.27 | Improved |
| add Style Swap (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 44.27 -> 45.84 | Improved |
| add Style Prompts (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 45.84 -> 46.16 | Improved |
| add Rare Class Generation (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 46.16 -> 46.47 | Improved |
| add Controlled Tiled MultiDiffusion within MRLF (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 45.19 -> 45.60 | Improved |
| add Latent Inpainting Diffusion to CTMD (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 45.60 -> 46.47 | Improved |
| add Style Swap and Style Prompting to generation (ablation) | DGInStyle generated data · FID to 5 real datasets (average) | FID: 136.68 -> 127.37 | Improved |
| increase generated images from 6000 to 8000 (ablation) | DGInStyle with DAFormer · GTA to 5 real datasets | mIoU (Avg5): 46.47 -> 46.39 | No clear effect |
| add DGInStyle data to HRDA with ResNet-101 | HRDA ResNet-101 · GTA to Dark Zurich | mIoU: 7.80 -> 16.16 | Improved |
Closest papers by meaning
- Diffusion Model as a Generalized Segmentation Learner (ECCV 2026 · similarity 0.89)
- Better to Teach than to Give: Domain Generalized Semantic Segmentation via Agent Queries with Diffusion Model Guidance (ICML 2025 · similarity 0.88)
- Style Nursing with Spatial and Semantic Guidance for Zero-Shot Traffic Scene Style Transfer (AAAI 2025 · similarity 0.87)
- Text-Image Alignment for Diffusion-Based Perception (CVPR 2024 · similarity 0.87)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · similarity 0.87)
- Stronger Fewer & Superior: Harnessing Vision Foundation Models for Domain Generalized Semantic Segmentation (CVPR 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.
