Research map / Label-efficient and robust learning / Domain adaptation and generalization
ACE: Adapting to Changing Environments for Semantic Segmentation
Zuxuan Wu, Xin Wang, Joseph E. Gonzalez, Tom Goldstein, Larry S. Davis
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 style replay memory (100 styles per task) (ablation) | ACE · Synthia Highway | mean mIoU: 61.2 -> 63.6 | Improved |
| add style replay memory (100 styles per task) (ablation) | ACE · Synthia NYC-like City | mean mIoU: 46.2 -> 48.9 | Improved |
| increase stored styles from 100 to 200 per task (ablation) | ACE · Synthia Highway | mean mIoU: 63.6 -> 64.0 | Improved |
| increase stored styles from 100 to 200 per task (ablation) | ACE · Synthia NYC-like City | mean mIoU: 48.9 -> 49.2 | Improved |
| adapt with ACE instead of source-only model | FCN-8s-ResNet101 · Synthia Highway | mean mIoU: 60.2 -> 63.6 | Improved |
| adapt with ACE instead of source-only model | FCN-8s-ResNet101 · Synthia NYC-like City | mean mIoU: 39.3 -> 48.9 | Improved |
| adapt with ACE instead of source-only model | DeepLab V3+ · Synthia NYC-like City | mean mIoU: 28.9 -> 50.4 | Improved |
| adapt with ACE instead of source-only model | DeepLab V3+ · Synthia Highway | mean mIoU: 51.7 -> 64.5 | Improved |
| adapt with ACE instead of source-only model | ResNet50-PSPNet · Synthia NYC-like City | mean mIoU: 48.4 -> 53.3 | Improved |
| transfer test images to source style (Source-Reverse) instead of adapting | source model · Synthia NYC-like City | mean mIoU: 39.3 -> 33.1 | Hurt |
| add feature-level ADDA alignment to ACE | ACE · Synthia Highway | mean mIoU: 63.6 -> 64.3 | Improved |
| add feature-level ADDA alignment to ACE | ACE · Synthia NYC-like City | mean mIoU: 48.9 -> 49.4 | Improved |
Closest papers by meaning
- Online Domain Adaptation for Semantic Segmentation in Ever-Changing Conditions (ECCV 2022 · similarity 0.89)
- On the Road to Online Adaptation for Semantic Image Segmentation (CVPR 2022 · similarity 0.89)
- Unsupervised Model Adaptation for Continual Semantic Segmentation (AAAI 2021 · similarity 0.88)
- Self-Supervised Augmentation Consistency for Adapting Semantic Segmentation (CVPR 2021 · similarity 0.88)
- Generalize Then Adapt: Source-Free Domain Adaptive Semantic Segmentation (ICCV 2021 · similarity 0.88)
- To Adapt or Not to Adapt? Real-Time Adaptation for Semantic Segmentation (ICCV 2023 · 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.
