Car Accident Video Collection For Detection Models

5,000 videos capture crashes occurring in traffic accidents. The dataset is designed to support traffic safety research and challenge detection models. It was created by ud-smart-city and last updated in April 2026.

Use Cases

• Train video object detection models to identify vehicles and crash events in traffic footage.
• Benchmark the performance of action recognition algorithms on crash sequences.
• Analyze pre-crash and post-crash vehicle dynamics from video frames for safety research.

Strengths

• Contains 5,000 videos, providing a substantial collection for model training.
• Specifically engineered to challenge detection models, indicating curated difficulty.

Limitations

• Unknown video resolution, frame rate, and duration limits usability assessment.
• Unknown geographic origin may limit model generalizability to specific regions.

Provenance

Source

ud-smart-city on Hugging Face.

Collection Method

null

Time Range

null

Freshness

Last updated in April 2026.

Geography

null