Research map / Label-efficient and robust learning / Domain adaptation and generalization
SAGE: Style-Adaptive Generalization for Privacy-Constrained Semantic Segmentation Across Domains
Qingmei Li, Yang Zhang, Peifeng Zhang, Haohuan Fu, Juepeng Zheng
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 prompt normalization (ablation) | SAGE APF · Cityscapes -> {B,M,G,S} avg | mIoU: 40.63 -> 41.85 | Improved |
| add softmax after prompt normalization (ablation) | SAGE APF · Cityscapes -> {B,M,G,S} avg | mIoU: 41.85 -> 42.17 | Improved |
| add tanh after normalization and softmax (ablation) | SAGE APF · Cityscapes -> {B,M,G,S} avg | mIoU: 42.17 -> 43.90 | Improved |
| add softmax alone without normalization (ablation) | SAGE APF · Cityscapes -> {B,M,G,S} avg | mIoU: 40.63 -> 40.49 | Hurt |
| add tanh alone (ablation) | SAGE APF · Cityscapes -> {B,M,G,S} avg | mIoU: 40.63 -> 41.59 | Improved |
| add modulation network to border prompt (A-Border vs Border) (ablation) | SAGE prompt generator · GTAV -> {C,B,M,S} avg | mIoU: 40.75 -> 42.09 | Improved |
| add modulation network to border prompt (A-Border vs Border) (ablation) | SAGE prompt generator · SYNTHIA -> {C,B,M,G} avg | mIoU: 36.84 -> 37.88 | Improved |
| add modulation to full-image prompt (A-Full vs Full) (ablation) | SAGE prompt generator · GTAV -> {C,B,M,S} avg | mIoU: 40.76 -> 40.80 | No clear effect |
| use border prompt instead of full-image prompt (adaptive) (ablation) | SAGE prompt generator · GTAV -> {C,B,M,S} avg | mIoU: 40.80 -> 42.09 | Improved |
| use meta initialization instead of zero init (ablation) | SAGE prompt template · GTAV -> {C,B,M,S} avg | mIoU: 40.76 -> 42.09 | Improved |
| use meta initialization instead of uniform init (ablation) | SAGE prompt template · SYNTHIA -> {C,B,M,G} avg | mIoU: 36.89 -> 37.88 | Improved |
| use four styles (L,P,V,Vo) instead of three (L,P,V) (ablation) | SAGE style set · Cityscapes -> {B,M,G,S} avg | mIoU: 38.26 -> 43.90 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- A2XP — A2XP: Towards Private Domain Generalization (CVPR 2024) · better on 3 of 3 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
- Exploring Semantic Consistency and Style Diversity for Domain Generalized Semantic Segmentation (AAAI 2025 · similarity 0.89)
- Exploiting Domain Properties in Language-Driven Domain Generalization for Semantic Segmentation (ICCV 2025 · similarity 0.88)
- Semantic-Aware Domain Generalized Segmentation (CVPR 2022 · similarity 0.88)
- A Simple Recipe for Language-guided Domain Generalized Segmentation (CVPR 2024 · similarity 0.87)
- FisherTune: Fisher-Guided Robust Tuning of Vision Foundation Models for Domain Generalized Segmentation (CVPR 2025 · similarity 0.87)
- Style Blind Domain Generalized Semantic Segmentation via Covariance Alignment and Semantic Consistence Contrastive Learning (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.
