Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
VZCrash is the largest publicly available dataset of real-world vehicle collision data, comprising over 31,000 validated crashes and 158,000 negative samples, all collected from a fleet of 73,010 commercial vehicles across the United States. The dataset provides high-frequency IMU telemetry, allowing for a comprehensive analysis of crash dynamics, and supports a benchmarking study of various crash detection methods, including heuristic and deep learning approaches. Key findings highlight that scaling data significantly enhances the performance of crash detection models, particularly in real-world applications, underscoring the critical role of data volume in model training.
Scaling data volume is crucial for training effective crash detection models, with implications for real-world deployment success.
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.