DriveMotion provides a large benchmark for predicting driver body movements
DriveMotion: A Large-Scale Multi-Source Benchmark for Driver Motion Sequence Modeling and Forecasting
Computer Vision and Pattern RecognitionHuman-Computer Interaction
Summary
Understanding how a driver moves their body can help predict what they might do next when driving. Existing datasets mostly focus on short clips or outside-the-car movements, but this work combines lots of data to continuously forecast driver motions from body keypoints inside the car. The authors also improve evaluation methods by focusing on motion around driving maneuvers, where body movements are more meaningful. Their models perform better than simple predictions, especially when trained on a larger combined dataset. This benchmark helps test future systems that predict driver behavior from body motions.
What this means in practice
- •For automotive safety engineers: Develop systems that predict driver actions by analyzing detailed body movements inside the car during vehicle maneuvers to improve accident prevention.
- •For vehicle telematics providers: Integrate continuous driver motion forecasting from multi-source data to enhance driver state monitoring and personalized alerts.
Authors
Yuhang Wang, Chuheng Wei, Jingxin Yang, Xishun Liao, Hao Zhou
Abstract
Driver motion can provide cues to ongoing behavior, attention, and near-term driving intent. However, most existing driver-centric datasets focus on recognizing predefined driver behaviors from short video clips, while human motion forecasting benchmarks largely target motion outside the vehicle. We introduce DriveMotion, a multi-source benchmark for continuous driver motion forecasting. DriveMotion contains 393 hours of 133-keypoint motion sequences at 10 Hz from 360 drivers, integrating naturalistic driving data, curated public in-cabin videos, and the AIDE dataset into a unified representation with per-joint validity masks and synchronized driving context. Naturalistic driving contains long periods of limited body movement, making uniformly sampled evaluation dominated by persistence and less sensitive to brief but behaviorally meaningful motion. To address this, we use dynamics-anchored evaluation, placing forecasting windows around vehicle maneuvers identified offline from CAN signals without providing CAN to the model at inference. Arm motion in pre-maneuver windows is 3.4x greater than in route-matched stable-driving controls. On these anchored windows, learned models reduce forecasting error over persistence by up to 15%, while maneuver-enriched training improves forecast-derived Part-State F1 by 44% over the zero-motion reference. Training on the full multi-source corpus further reduces forecasting error on held-out web drivers by 38% compared with BATON-only training. DriveMotion provides identity-disjoint splits, fixed evaluation subsets, and reference implementations for reproducible evaluation of continuous driver motion forecasting. The dataset and benchmark are available at https://huggingface.co/datasets/HenryYHW/DriveMotion