VZCrash: A Large-Scale IMU Dataset of Ego-Vehicle Crashes
Provides a large-scale benchmark for crash detection using IMU data, enabling researchers to develop and evaluate models for commercial vehicle safety.
VZCrash is the largest public IMU dataset of real-world vehicle crashes, containing over 31,000 validated crashes and 158,000 negative samples. Benchmarking shows that scaling data significantly improves crash detection model performance, especially for real-world deployment.
We introduce VZCrash, the largest publicly available dataset of real-world vehicle collision data featuring Inertial Measurement Unit (IMU) telemetry. The dataset contains more than 31,000 validated crashes and 158,000 negative samples, including hard cases and distractors. Each sample includes acceleration and angular velocity at 100 Hz, and GPS speed at 1 Hz. Events in VZCrash were captured by devices installed on a fleet of 73,010 commercial vehicles of different sizes driving in the United States over the span of several years. We also present an extensive experimental study enabled by the volume of the dataset. We first benchmark several different approaches, from a simple threshold-based heuristic to state-of-the-art deep learning models. Then, we present an experiment demonstrating the importance of scaling data to train high-quality crash detection models, and we show that scale is especially important when these models need to be deployed into a real-world environment.