Research map / Video / Video segmentation and tracking
EventDrive: Event Cameras for Vision-Language Driving Intelligence
Dongyue Lu, Rong Li, Ao Liang, Lingdong Kong, Wei Yin, Lai Xing Ng, Benoit R. Cottereau, Camille Simon Chane, Wei Tsang Ooi
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 event stream to frame input | Qwen2.5-VL-7B fine-tuned · EventDrive Planning | L2 error: 4.54 -> 3.66 | Improved |
| add event stream to frame input | Qwen2.5-VL-7B fine-tuned · EventDrive grounding | mIoU: 67.19 -> 72.56 | Improved |
| add event stream to frame input | Qwen2.5-VL-7B fine-tuned · EventDrive Prediction speed | accuracy: 36.84 -> 42.44 | Improved |
| add event stream to frame input | Qwen2.5-VL-7B fine-tuned · EventDrive Understanding | QA accuracy: 58.44 -> 65.46 | Improved |
| fine-tune on EventDrive instead of zero-shot | Qwen2.5-VL-7B-Instruct · EventDrive Planning | L2 error: 8.78 -> 4.54 | Improved |
| use event-only input instead of EventGPT-7B zero-shot | EventDrive-VLM · EventDrive Planning | L2 error: 11.42 -> 6.89 | Improved |
| increase horizons from 1 to 3 (ablation) | EventDrive-VLM · EventDrive perception | QA accuracy: 82.40 -> 83.66 | Improved |
| increase horizons from 3 to 5 (ablation) | EventDrive-VLM · EventDrive perception | QA accuracy: 83.66 -> 83.95 | No clear effect |
| use MoE top-1 expert instead of Add aggregation (ablation) | EventDrive-VLM · EventDrive perception | QA accuracy: 76.76 -> 83.66 | Improved |
| use MoE top-1 expert instead of weighted sum (ablation) | EventDrive-VLM · EventDrive prediction | accuracy: 61.08 -> 64.96 | Improved |
| use Event Q-Former instead of concatenation (ablation) | EventDrive-VLM · EventDrive perception | QA accuracy: 79.35 -> 83.66 | Improved |
| use Event Q-Former instead of cross-attention (ablation) | EventDrive-VLM · EventDrive planning | L2 error: 3.69 -> 3.66 | Mixed |
Closest papers by meaning
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.89)
- Talk2Event: Grounded Understanding of Dynamic Scenes from Event Cameras (NeurIPS 2025 · similarity 0.89)
- SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving (CVPR 2026 · similarity 0.88)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- STSBench: A Spatio-temporal Scenario Benchmark for Multi-modal Large Language Models in Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- RE-VLM: Event-Augmented Vision-Language Model for Scene Understanding (CVPR 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.
