Research map / Label-efficient and robust learning / Domain adaptation and generalization
SHIFT: A Synthetic Driving Dataset for Continuous Multi-Task Domain Adaptation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Test a model trained on clear-daytime under shifted weather and lighting | SHIFT perception models · SHIFT discrete domain shifts | Perception performance: 83.6 -> 42.8 | Hurt |
| Adapt incrementally through intermediate domains instead of adapting directly | DRN-D-54 semantic segmentation · Daytime → night | mIoU: 45.3 -> 47.3 | Helped |
| Adapt incrementally through intermediate domains | DRN-D-54 semantic segmentation · Clear → rainy | mIoU: 61.0 -> 64.9 | Helped |
| Apply domain adaptation to the foggy target | DRN-D-54 semantic segmentation · Clear → foggy | mIoU: 62.4 -> 59.1; 62.4 -> 57.3 | Hurt |
| Use continuous test-time adaptation when only a small amount of target data is available | Semantic segmentation adaptation · Daytime → night, limited sequences | mIoU | Helped |
| Use incremental adaptation when enough sequences are available | Semantic segmentation adaptation · Daytime → night, increasing sequence count | mIoU | Helped |
| Adapt frame by frame through a continuous shift | TENT semantic segmentation · Continuous shifts with return to source | Source-domain performance | Hurt |
| Jointly train semantic segmentation, depth, and instance segmentation (ablation) | DRN-D-54 multitask model · SHIFT, semantic segmentation OOD average | mIoU (%): 24.9 -> 40.4 | Helped |
| Jointly train semantic segmentation, depth, and instance segmentation (ablation) | DRN-D-54 multitask model · SHIFT, depth OOD average | SILog: 49.6 -> 37.3 | Helped |
| Jointly train semantic segmentation, depth, and instance segmentation (ablation) | DRN-D-54 multitask model · SHIFT, instance segmentation OOD average | mAP (%): 34.2 -> 36.4 | Helped |
| Jointly train optical flow with semantic segmentation (ablation) | DRN-D-54 multitask model · SHIFT, optical flow OOD average | EPE (px): 8.6 -> 10.4 | Hurt |
| Jointly train depth and semantic segmentation (ablation) | DRN-D-54 multitask model · SHIFT, depth OOD average | SILog: 49.6 -> 43.9 | Helped |
Closest papers by meaning
- Towards Optimal Strategies for Training Self-Driving Perception Models in Simulation (NeurIPS 2021 · similarity 0.89)
- The Road Ahead in Autonomous Driving: The KITScenes Multimodal Dataset (NeurIPS 2026 · similarity 0.87)
- Panacea: Panoramic and Controllable Video Generation for Autonomous Driving (CVPR 2024 · similarity 0.86)
- SceneCrafter: Controllable Multi-View Driving Scene Editing (CVPR 2025 · similarity 0.85)
- DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation (CVPR 2025 · similarity 0.85)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.85)
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.