Research map / Label-efficient and robust learning / Domain adaptation and generalization
Multi-Target Adversarial Frameworks for Domain Adaptation in Semantic Segmentation
Antoine Saporta, Tuan-Hung Vu, Matthieu Cord, Patrick Pérez
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 MTKT multi-teacher/single-student distillation instead of merging targets | Multi-Target Baseline · GTA5->Cityscapes+Mapillary | mIoU Avg.: +2.0% | Improved |
| use multi-discriminator alignment instead of merging targets | Multi-Target Baseline · GTA5->Cityscapes+Mapillary | mIoU Avg.: +0.6% | Improved |
| use multi-discriminator instead of merged-target training | Multi-Target Baseline · GTA5->Cityscapes+IDD | mIoU Avg.: 67.4 -> 67.3 | No clear effect |
| use MTKT instead of merged-target training | Multi-Target Baseline · GTA5->Cityscapes+Mapillary+IDD | mIoU Avg.: 67.8 -> 69.1 | Improved |
| use MTKT instead of merged-target training | Multi-Target Baseline · Cityscapes->Mapillary+IDD | mIoU Avg.: 67.0 -> 69.0 | Improved |
| train on multiple target domains instead of a single target | AdvEnt single-target · GTA5->Cityscapes+Mapillary | mIoU Avg.: 66.6 -> 68.9 | Improved |
| add adversarial training on the target-agnostic branch of MTKT (ablation) | MTKT · GTA5->Cityscapes+Mapillary | mIoU Avg.: -0.9% | Hurt |
| add pseudo-labeling (ESL) to MTKT, refining target-specific head only (ablation) | MTKT · GTA5->Cityscapes+IDD | mIoU Avg.: 68.2 -> 69.8 | Improved |
| add pseudo-labeling (ESL) to multi-target baseline (ablation) | Multi-Target Baseline · GTA5->Cityscapes+IDD | mIoU Avg.: 67.4 -> 68.9 | Improved |
| refine target-agnostic model too with pseudo-labels (PL 3) rather than head only (PL 1) (ablation) | MTKT + PL · GTA5->Cityscapes+IDD | mIoU Avg.: 69.8 -> 69.9 | No clear effect |
| use MTKT instead of multi-target baseline for direct transfer to unseen domain | MTKT · Direct transfer to Mapillary | mIoU: 66.7 -> 70.6 | Improved |
| use multi-discriminator instead of multi-target baseline for direct transfer to unseen domain | Multi-Dis. · Direct transfer to Mapillary | mIoU: 66.7 -> 65.9 | Hurt |
Closest papers by meaning
- Focus on Your Target: A Dual Teacher-Student Framework for Domain-Adaptive Semantic Segmentation (ICCV 2023 · similarity 0.93)
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.93)
- Multi-Target Domain Adaptation With Collaborative Consistency Learning (CVPR 2021 · similarity 0.93)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.92)
- Balanced Learning for Domain Adaptive Semantic Segmentation (ICML 2025 · similarity 0.91)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · similarity 0.90)
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.
