Research map / Label-efficient and robust learning / Domain adaptation and generalization
Towards Optimal Strategies for Training Self-Driving Perception Models in Simulation
Read the paperCheck 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 |
|---|---|---|---|
| Use the proposed f-DAL Pearson adaptation with pseudo-labels instead of RS-No Adaptation | Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 9.76 -> 17.84 | Helped |
| Use the proposed f-DAL Pearson adaptation with pseudo-labels instead of the strongest listed adaptation baseline | Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 13.76 -> 17.84 | Helped |
| Use the proposed adaptation method instead of RS-No Adaptation | Point Pillars · nuScenes val, BEV vehicle segmentation | IOU: 15.09 -> 17.20 | Helped |
| Use the proposed adaptation method instead of DANN | Point Pillars · nuScenes val, BEV vehicle segmentation | IOU: 14.41 -> 17.20 | Helped |
| Add domain adaptation and pseudo-label training (ablation) | Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 10.33 -> 17.84 | Helped |
| Use latent-space f-DAL Pearson adaptation with pseudo-labels instead of MUNIT style transfer (ablation) | Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 14.32 -> 17.84 | Helped |
| Use f-DAL Pearson adaptation with pseudo-labels instead of DANN (ablation) | Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 14.45 -> 17.84 | Helped |
| Use Pearson rather than Jensen discrepancy and add pseudo-labels (ablation) | f-DAL on Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 16.19 -> 17.84 | Helped |
| Add pseudo-labels to f-DAL Pearson (ablation) | Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU: 17.03 -> 17.84 | Helped |
| Sample NPC locations using a spatial prior to reduce the label-marginal mismatch rather than sampling from town road geometry | CARLA-trained Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU | Helped |
| Use data with lower source-target label-marginal JSD | CARLA-trained models · nuScenes val, BEV vehicle segmentation | IOU | Helped |
| Vary weather while keeping the NPC sampling strategy in the town (ablation) | CARLA-trained Lift-Splat · nuScenes val, BEV vehicle segmentation | IOU | No clear effect |
Closest papers by meaning
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- Transfer Your Perspective: Controllable 3D Generation from Any Viewpoint in a Driving Scene (CVPR 2025 · similarity 0.88)
- The Road Ahead in Autonomous Driving: The KITScenes Multimodal Dataset (NeurIPS 2026 · similarity 0.88)
- Uncertainty-Guided Never-Ending Learning to Drive (CVPR 2024 · similarity 0.87)
- ReSimAD: Zero-Shot 3D Domain Transfer for Autonomous Driving with Source Reconstruction and Target Simulation (ICLR 2024 · similarity 0.87)
- JiSAM: Alleviate Labeling Burden and Corner Case Problems in Autonomous Driving via Minimal Real-World Data (CVPR 2025 · 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.