Research map / Label-efficient and robust learning / Domain adaptation and generalization
Label-Driven Reconstruction for Domain Adaptation in Semantic Segmentation
Jinyu Yang, Weizhi An, Sheng Wang, Xinliang Zhu, Chaochao Yan, Junzhou Huang
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 |
|---|---|---|---|
| replace source-to-target with target-to-source translation (ablation) | DeepLab-ResNet101 baseline · GTA5 to Cityscapes | mIoU: 48.5 -> 49.1 | Improved |
| replace source-to-target with target-to-source translation (ablation) | FCN-VGG16 baseline · GTA5 to Cityscapes | mIoU: 41.3 -> 42.3 | Improved |
| replace source-to-target with target-to-source translation (ablation) | DeepLab-ResNet101 baseline · SYNTHIA to Cityscapes | mIoU: 51.4 -> 52.0 | Improved |
| replace source-to-target with target-to-source translation (ablation) | FCN-VGG16 baseline · SYNTHIA to Cityscapes | mIoU: 39.0 -> 40.1 | Improved |
| add label-driven reconstruction network (ablation) | target-to-source translation, ResNet101 · GTA5 to Cityscapes | mIoU: 49.1 -> 49.5 | Improved |
| add label-driven reconstruction network (ablation) | target-to-source translation, VGG16 · GTA5 to Cityscapes | mIoU: 42.3 -> 43.6 | Improved |
| add label-driven reconstruction network (ablation) | target-to-source translation, ResNet101 · SYNTHIA to Cityscapes | mIoU: 52.0 -> 53.1 | Improved |
| add label-driven reconstruction network (ablation) | target-to-source translation, VGG16 · SYNTHIA to Cityscapes | mIoU: 40.1 -> 41.1 | Improved |
| reconstruct from label space instead of feature space (ablation) | reconstruction network, VGG16 · GTA5 to Cityscapes | mIoU: 41.48 -> 43.6 | Improved |
| reconstruct from label space instead of feature space (ablation) | reconstruction network, VGG16 · SYNTHIA to Cityscapes | mIoU: 40.13 -> 41.1 | Improved |
| set Boltzmann temperature tau to 0.001 instead of 1 (ablation) | reconstruction from label space, VGG16 · GTA5 to Cityscapes | mIoU: 41.5 -> 43.6 | Improved |
| add VGG perceptual loss to reconstruction loss (ablation) | reconstruction loss, VGG16 · GTA5 to Cityscapes | mIoU: 41.53 -> 42.82 | Improved |
Closest papers by meaning
- CrDoCo: Pixel-Level Domain Transfer With Cross-Domain Consistency (CVPR 2019 · similarity 0.92)
- Multi-Source Domain Adaptation With Collaborative Learning for Semantic Segmentation (CVPR 2021 · similarity 0.91)
- Domain transfer through deep activation matching (ECCV 2018 · similarity 0.91)
- An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation (AAAI 2020 · similarity 0.91)
- Consistency Regularization with High-dimensional Non-adversarial Source-guided Perturbation for Unsupervised Domain Adaptation in Segmentation (AAAI 2021 · similarity 0.91)
- Taking a Closer Look at Domain Shift: Category-Level Adversaries for Semantics Consistent Domain Adaptation (CVPR 2019 · similarity 0.91)
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.
