Research map / Label-efficient and robust learning / Domain adaptation and generalization
SePiCo: Semantic-Guided Pixel Contrast for Domain Adaptive Semantic Segmentation
Binhui Xie, Shuang Li, Mingjia Li, Chi Harold Liu, Gao Ming Huang, Guoren Wang
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 distribution-aware pixel contrast (DistCL) instead of centroid-aware ProtoCL | SePiCo · GTAV to Cityscapes | mIoU: 58.8 -> 61.0 | Improved |
| use DistCL instead of memory-bank BankCL | SePiCo · GTAV to Cityscapes | mIoU: 59.8 -> 61.0 | Improved |
| apply SePiCo (DistCL) self-training with pixel contrast over source-only training | DeepLab-V2 ResNet-101 · GTAV to Cityscapes | mIoU: +22.4% mIoU | Improved |
| apply SePiCo (DistCL) over source-only training | DeepLab-V2 ResNet-101 · SYNTHIA to Cityscapes (16 classes) | mIoU: +24.6% mIoU | Improved |
| use SePiCo (DistCL) one-stage training instead of multi-stage ProDA | DeepLab-V2 ResNet-101 · GTAV to Cityscapes | mIoU: +3.5% mIoU | Improved |
| use DistCL instead of ProtoCL or BankCL | SePiCo · SYNTHIA to Cityscapes (16 classes) | mIoU: 56.8 -> 58.1 | Improved |
| use DistCL instead of ProtoCL | SePiCo · Cityscapes to Dark Zurich | mIoU: 42.6 -> 45.4 | Improved |
| replace DeepLab-V2 with DAFormer architecture and add DistCL | SePiCo · GTAV to Cityscapes (DAFormer) | mIoU: 68.3 -> 70.3 | Improved |
| add DistCL to DAFormer | SePiCo · Cityscapes to Dark Zurich (DAFormer) | mIoU: 48.5 -> 54.2 | Improved |
| add ProtoCL to DAFormer | SePiCo (ProtoCL) · GTAV to Cityscapes (DAFormer) | mIoU: 68.3 -> 68.5 | No clear effect |
| add DistCL to DAFormer, test on unseen night domains | SePiCo · Nighttime Driving (generalization from Dark Zurich) | mIoU: 51.8 -> 56.9 | Improved |
| add DistCL to DAFormer, test on unseen night domains | SePiCo · BDD100k-night (generalization from Dark Zurich) | mIoU: 33.9 -> 40.6 | 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.
- DAFormer — DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation (CVPR 2022) · better on 4 of 4 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
- Bi-directional Contrastive Learning for Domain Adaptive Semantic Segmentation (ECCV 2022 · similarity 0.92)
- TACS: Taxonomy Adaptive Cross-Domain Semantic Segmentation (ECCV 2022 · similarity 0.90)
- Cross-Image Pixel Contrasting for Semantic Segmentation (TPAMI 2024 · similarity 0.90)
- Exploring Cross-Image Pixel Contrast for Semantic Segmentation (ICCV 2021 · similarity 0.90)
- Pixel-Level Cycle Association: A New Perspective for Domain Adaptive Semantic Segmentation (NeurIPS 2020 · similarity 0.90)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · 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.
