Research map / Label-efficient and robust learning / Domain adaptation and generalization
Self-Supervised CycleGAN for Object-Preserving Image-to-Image Domain Adaptation
Xinpeng Xie, Jiawei Chen, Yuexiang Li, Linlin Shen, Kai Ma, Yefeng Zheng
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 content registration branch (content consistency loss) to CycleGAN (ablation) | CycleGAN · CamVid cloudy-to-sunny (cloudy test) | mIoU: 26.26 -> 45.63 | Improved |
| add domain classification branch to CycleGAN (ablation) | CycleGAN · CamVid cloudy-to-sunny (cloudy test) | mIoU: 26.26 -> 45.86 | Improved |
| combine content registration and domain classification (ablation) | CycleGAN · CamVid cloudy-to-sunny (cloudy test) | mIoU: 26.26 -> 51.40 | Improved |
| add domain classification on top of content registration (ablation) | OP-GAN · CamVid cloudy-to-sunny (cloudy test) | mIoU: 45.63 -> 51.40 | Improved |
| add content registration on top of domain classification (ablation) | OP-GAN · CamVid cloudy-to-sunny (cloudy test) | mIoU: 45.86 -> 51.40 | Improved |
| apply OP-GAN translation instead of direct transfer | PSPNet · CamVid cloudy-to-sunny (cloudy test) | mIoU: 41.86 -> 51.40 | Improved |
| apply CycleGAN translation instead of direct transfer | PSPNet · CamVid cloudy-to-sunny (cloudy test) | mIoU: 41.86 -> 26.26 | Hurt |
| apply UNIT translation instead of direct transfer | PSPNet · CamVid cloudy-to-sunny (cloudy test) | mIoU: 41.86 -> 19.94 | Hurt |
| apply DRIT translation instead of direct transfer | PSPNet · CamVid cloudy-to-sunny (cloudy test) | mIoU: 41.86 -> 18.20 | Hurt |
| apply OP-GAN translation instead of direct transfer | PSPNet · SYNTHIA night-to-day (night test) | mIoU: 44.49 -> 50.86 | Improved |
| apply CycleGAN translation instead of direct transfer | PSPNet · SYNTHIA night-to-day (night test) | mIoU: 44.49 -> 22.88 | Hurt |
| apply UNIT translation instead of direct transfer | PSPNet · SYNTHIA night-to-day (night test) | mIoU: 44.49 -> 34.97 | Hurt |
Closest papers by meaning
- AugGAN: Cross Domain Adaptation with GAN-based Data Augmentation (ECCV 2018 · similarity 0.88)
- TraVeLGAN: Image-To-Image Translation by Transformation Vector Learning (CVPR 2019 · similarity 0.88)
- Learning From Synthetic Data: Addressing Domain Shift for Semantic Segmentation (CVPR 2018 · similarity 0.87)
- Self-Supervised Dense Consistency Regularization for Image-to-Image Translation (CVPR 2022 · similarity 0.87)
- Conditional Generative Adversarial Network for Structured Domain Adaptation (CVPR 2018 · similarity 0.87)
- Breaking the Cycle - Colleagues Are All You Need (CVPR 2020 · similarity 0.87)
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.
