DroneDetectionDataset

Overview

A real-world object detection dataset specifically designed for detecting quadcopter UAVs. It contains over 50,000 training images and 5,000 test images with annotations in PASCAL VOC format.

Key Statistics

• 51,446 training images
• 5,375 test images
• Single class: "drone" (quadcopter UAV)
• Various lighting conditions and environments
• Different distances and angles

Strengths

• Large-scale, focused dataset
• Real-world capture conditions
• Well-structured train/test split
• Standard annotation format

Limitations

• Limited to quadcopter UAVs
• Single class only
• No temporal information
• Limited environmental diversity

Applications

• Counter-drone systems
• Drone detection for security
• Airspace monitoring
• Restricted area surveillance

Source

GitHub: https://github.com/Maciullo/DroneDetectionDataset