Research map / Label-efficient and robust learning / Domain adaptation and generalization
Condition-Invariant Semantic Segmentation
Christos Sakaridis, David Brüggemann, Fisher Yu, Luc Van Gool
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 feature invariance loss on source views (ablation) | HRDA with FDA stylized CE · ACDC val ablation (Cityscapes->ACDC) | mIoU: 65.7 -> 66.6 | Improved |
| add feature invariance loss on target views to plain HRDA (ablation) | HRDA · ACDC val ablation (Cityscapes->ACDC) | mIoU: 64.1 -> 66.9 | Improved |
| add both source and target feature invariance losses (ablation) | HRDA · ACDC val ablation (Cityscapes->ACDC) | mIoU: 64.1 -> 68.2 | Improved |
| replace original CE with FDA stylized-source CE (ablation) | HRDA · ACDC val ablation (Cityscapes->ACDC) | mIoU: 64.1 -> 65.7 | Improved |
| add CE on both original and stylized source images (ablation) | HRDA with FDA · ACDC val ablation (Cityscapes->ACDC) | mIoU: 65.7 -> 65.1 | No clear effect |
| add CE on stylized target images with source invariance (ablation) | CISS source-only · ACDC val ablation (Cityscapes->ACDC) | mIoU: 66.6 -> 65.7 | Hurt |
| add CE on stylized images on top of CISS (ablation) | CISS · ACDC val ablation (Cityscapes->ACDC) | mIoU: 68.2 -> 68.0 | No clear effect |
| apply invariance on internal encoder features instead of network outputs (ablation) | CISS-source · ACDC val ablation (Cityscapes->ACDC) | mIoU | Improved |
| use squared Frobenius norm instead of L1 norm for invariance loss (ablation) | CISS-source · ACDC val ablation (Cityscapes->ACDC) | mIoU | Improved |
| vary source weight lambda_s from 50 to 1000 (ablation) | CISS-source · ACDC val ablation (Cityscapes->ACDC) | mIoU: 65.8, 65.6, 66.6, 65.7, 65.9 | No clear effect |
| vary target weight lambda_t from 20 to 500 (ablation) | CISS-target · ACDC val ablation (Cityscapes->ACDC) | mIoU: 66.7, 66.6, 66.9, 66.1, 65.2 | Mixed |
| add feature invariance loss to HRDA | HRDA · Cityscapes->Dark Zurich | mIoU: 55.9 -> 60.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- HRDA — HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation (ECCV 2022) · better on 8 of 9 shared results
- DAFormer — DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation (CVPR 2022) · better on 2 of 2 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
- Contrastive Model Adaptation for Cross-Condition Robustness in Semantic Segmentation (ICCV 2023 · similarity 0.90)
- Domain Stylization: A Fast Covariance Matching Framework Towards Domain Adaptation (TPAMI 2020 · similarity 0.89)
- A Curriculum Domain Adaptation Approach to the Semantic Segmentation of Urban Scenes (TPAMI 2019 · similarity 0.89)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.88)
- Weakly Supervised Semantic Segmentation for Driving Scenes (AAAI 2024 · similarity 0.88)
- MRFP: Learning Generalizable Semantic Segmentation from Sim-2-Real with Multi-Resolution Feature Perturbation (CVPR 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.
