Research map / Label-efficient and robust learning / Domain adaptation and generalization
Dual Path Learning for Domain Adaptation of Semantic Segmentation
Yiting Cheng, Fangyun Wei, Jianmin Bao, Dong Chen, Fang Wen, Wenqiang Zhang
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 dual perceptual loss (DPIT) instead of single path perceptual loss (ablation) | image translation, M_T mIoU · GTA5->Cityscapes ablation (first iteration) | mIoU: 51.1 -> 51.8 | Improved |
| add perceptual loss to CycleGAN translation (SPIT) (ablation) | image translation, M_T mIoU · GTA5->Cityscapes ablation (first iteration) | mIoU: 48.5 -> 51.1 | Improved |
| use dual path pseudo labels (weighted) instead of single path (ablation) | pseudo label generation, M_T mIoU · GTA5->Cityscapes ablation (first iteration) | mIoU: 50.0 -> 51.8 | Improved |
| weighted averaging instead of max of two paths (ablation) | DPPLG, M_T mIoU · GTA5->Cityscapes ablation (first iteration) | mIoU: 50.6 -> 51.8 | Improved |
| weighted averaging instead of intersecting pseudo labels (ablation) | DPPLG, M_T mIoU · GTA5->Cityscapes ablation (first iteration) | mIoU: 50.3 -> 51.8 | Improved |
| add label correction versus raw ground-truth labels for M_T warm-up (ablation) | M_T warm-up · GTA5->Cityscapes (warm-up) | mIoU: 46.2 -> 48.5 | Improved |
| add label correction versus raw pseudo labels for M_T warm-up (ablation) | M_T warm-up · GTA5->Cityscapes (warm-up) | mIoU: 44.3 -> 48.5 | Improved |
| add M_T warm-up strategy (ablation) | DPL ResNet101 · GTA5->Cityscapes val | mIoU: 49.0 -> 52.8 | Improved |
| run first DPAS iteration after warm-up (ablation) | M_T · GTA5->Cityscapes ablation (first iteration) | mIoU: 48.5 -> 51.8 | Improved |
| separate segmentation models instead of sharing all weights (ablation) | M_S and M_T · GTA5->Cityscapes ablation (first iteration) | mIoU: 49.2 -> 51.8 | Improved |
| separate segmentation models instead of sharing first 4 blocks (ablation) | M_S and M_T · GTA5->Cityscapes ablation (first iteration) | mIoU: 49.6 -> 51.8 | Improved |
| use alpha 0.5 instead of only target path (alpha 1) (ablation) | pseudo label weighting · GTA5->Cityscapes ablation (first iteration) | mIoU: 50.0 -> 51.8 | Improved |
Closest papers by meaning
- ADPL: Adaptive Dual Path Learning for Domain Adaptation of Semantic Segmentation (TPAMI 2023 · similarity 0.96)
- Semi-Supervised Domain Adaptation Based on Dual-Level Domain Mixing for Semantic Segmentation (CVPR 2021 · similarity 0.90)
- Learning to Adapt via Latent Domains for Adaptive Semantic Segmentation (NeurIPS 2021 · similarity 0.90)
- SemiDAViL: Semi-supervised Domain Adaptation with Vision-Language Guidance for Semantic Segmentation (CVPR 2025 · similarity 0.89)
- ADAS: A Direct Adaptation Strategy for Multi-Target Domain Adaptive Semantic Segmentation (CVPR 2022 · similarity 0.89)
- Universal Domain Adaptation 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.
