Research map / Label-efficient and robust learning / Noisy labels
Crash to Not Crash: Learn to Identify Dangerous Vehicles Using a Simulator
Hoon Kim, Kangwook Lee, Gyeongjo Hwang, Changho Suh
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 |
|---|---|---|---|
| train on synthetic GTACrash instead of real YouTubeCrash | CNN (RGB + B.Box) · YouTubeCrash test | AUC: 0.8730 -> 0.9086 | Improved |
| train on GTACrash instead of YouTubeCrash | 5-layer NN (B.Box-size) · YouTubeCrash test | AUC: 0.8766 -> 0.8812 | Improved |
| add RGB channels to bounding box input (RGB+B.Box vs B.Box), GTACrash (ablation) | CNN · YouTubeCrash test | AUC: 0.8992 -> 0.9086 | Improved |
| use RGB+B.Box complex model when trained on small real data | CNN (RGB + B.Box) · YouTubeCrash test | AUC: 0.8766 -> 0.8730 | Hurt |
| add label adaptation via CTRA path prediction (ablation) | CNN (RGB + B.Box) on GTACrash · YouTubeCrash test | AUC: 0.9086 -> 0.9154 | Improved |
| add feature adaptation (backward CycleGAN mapping) (ablation) | CNN (RGB + B.Box) on GTACrash · YouTubeCrash test | AUC: 0.9086 -> 0.9102 | Improved |
| add both label and feature adaptation (ablation) | CNN (RGB + B.Box) on GTACrash · YouTubeCrash test | AUC: 0.9086 -> 0.9164 | Improved |
| add adaptation with ResNet50 backbone (ablation) | CNN (RGB + B.Box) ResNet50 · YouTubeCrash test | AUC: 0.9097 -> 0.9241 | Improved |
| use GTACrash-trained model vs real-trained at TTC 1.6-1.8 s | CNN (RGB + B.Box) · YouTubeCrash test | missed detection rate: -18.5% | Improved |
| use GTACrash-trained complex model at very short TTC (<0.8 s) | CNN (RGB + B.Box) · YouTubeCrash test | missed detection rate | Hurt |
| use all three frames per sample (ablation) | CNN (RGB + B.Box) · YouTubeCrash test | AUC | Improved |
Closest papers by meaning
- Progressive Risk Estimation for Accident Anticipation (NeurIPS 2026 · similarity 0.87)
- DeepAccident: A Motion and Accident Prediction Benchmark for V2X Autonomous Driving (AAAI 2024 · similarity 0.86)
- DROID: Driver-Centric Risk Object IDentification (TPAMI 2023 · similarity 0.85)
- Accident Anticipation via Temporal Occurrence Prediction (NeurIPS 2025 · similarity 0.85)
- DiffScene: Diffusion-Based Safety-Critical Scenario Generation for Autonomous Vehicles (AAAI 2025 · similarity 0.83)
- Generating Traffic Scenarios via In-Context Learning to Learn Better Motion Planner (AAAI 2025 · similarity 0.83)
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.
