VZCrash: Telemetry of Ego-Vehicle Crashes with 31,000 Verified Events

Almost 190,000 unique events of vehicle telemetry, including over 31,000 verified crashes. The VZCrash dataset contains 100 Hz tri-axial accelerometer and gyroscope data alongside 1 Hz GPS-derived speed. It was created by vzc-research-chapter to accompany a paper accepted for the 2026 IEEE International Conference.

Use Cases

• Training crash detection models based on high-frequency accelerometer and gyroscope signals.
• Analyzing pre-crash vehicle dynamics using the 100 Hz inertial measurement unit (IMU) data.
• Correlating crash severity with GPS-derived speed data captured at 1 Hz.
• Benchmarking anomaly detection algorithms for automotive safety applications.

Strengths

• Contains over 31,000 verified positive crash events.
• Provides high-frequency (100 Hz) IMU data for detailed temporal analysis.
• Includes nearly 190,000 unique events for model training and validation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the full dataset is unknown, which may limit suitability assessment.
• Freshness should be verified as the last metadata update was 2026-05-14.

Provenance

Source

huggingface dataset by vzc-research-chapter

Collection Method

Telemetry data collection from ego-vehicles, likely via onboard sensors.

Freshness

Last updated 2026-05-14 10:01:10