Research map / Label-efficient and robust learning / Recognition and long-tail
Real-to-Virtual Domain Unification for End-to-End Autonomous Driving
Luona Yang, Xiaodan Liang, Tairui Wang, Eric Xing
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 |
|---|---|---|---|
| transform real images to virtual domain with DU-Drive(single) instead of fine-tuning on Carla data | PilotNet · Udacity | steering MAE: 6.053 -> 3.571 | Improved |
| transform real images to virtual domain with DU-Drive(single) instead of fine-tuning on TORCS data | PilotNet · Udacity | steering MAE: 5.808 -> 4.558 | Improved |
| transform real images to virtual domain with DU-Drive(single) instead of fine-tuning on TORCS data | PilotNet · Comma.ai | steering MAE: 1.203 -> 1.061 | Improved |
| add joint training of prediction and transformation to conditional GAN (ablation) | cGAN · Udacity | steering MAE: 5.921 -> 4.558 | Improved |
| train one predictor on both real datasets directly without domain unification | PilotNet · Comma.ai | steering MAE: 1.208 -> 5.988 | Hurt |
| use unified model over two real domains instead of single-domain model | DU-Drive · Comma.ai | steering MAE: 1.061 -> 1.079 | No clear effect |
| add domain unification with 20% labeled Udacity data (Carla) | PilotNet · Udacity semi-supervised | steering MAE: 7.86 -> 6.02 | Improved |
| add domain unification with 50% labeled Udacity data (Carla) | PilotNet · Udacity semi-supervised | steering MAE: 7.11 -> 5.15 | Improved |
| use unified instead of single model with 100% data (Carla) | DU-Drive · Udacity semi-supervised | steering MAE: 3.57 -> 3.81 | Hurt |
| use generated virtual images instead of real images (Carla, Udacity) | input images · Udacity | image variance: 82745 -> 31650 | Improved |
Closest papers by meaning
- DualAD: Disentangling the Dynamic and Static World for End-to-End Driving (CVPR 2024 · similarity 0.87)
- VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.87)
- UniDrive: Towards Universal Driving Perception Across Camera Configurations (ICLR 2025 · similarity 0.87)
- ReSimAD: Zero-Shot 3D Domain Transfer for Autonomous Driving with Source Reconstruction and Target Simulation (ICLR 2024 · similarity 0.87)
- CoC-VLA: Delving into Adversarial Domain Transfer for Explainable Autonomous Driving via Chain-of-Causality Visual-Language-Action Model (NeurIPS 2025 · similarity 0.86)
- DriveHierarchy: A Benchmark for Diagnosing VLM Driving Capabilities from Open-Loop Understanding to Closed-Loop Execution (NeurIPS 2026 · similarity 0.86)
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.
