YouTube-GDD: Gun Detection with Rich Contextual Information

YouTube-GDD is a gun detection dataset sourced from YouTube videos, created by UCAS-GYX and updated in December 2025. It provides visual data for identifying firearms within diverse contextual environments to support deep learning and object detection research.

Use Cases

• Training YOLO or Faster R-CNN models to detect firearms in low-resolution or motion-blurred video frames
• Evaluating the impact of background clutter on gun detection precision using the contextual labels
• Developing automated safety monitoring systems for video platforms using the provided gun annotations

Strengths

• Released under the permissive Apache-2.0 license
• Sourced from real-world YouTube footage providing high contextual variety
• Specifically targets 'challenging' detection scenarios

Limitations

• Unknown record count and file size
• Potential for label noise inherent in scraped video data
• Lack of detailed documentation regarding specific firearm categories

Provenance

Source

UCAS-GYX via GitHub

Collection Method

scraped

Freshness

Last updated December 2025.

Geography

global