Real-Time Target Detection for UAVs using Finetuned YOLOv3 and Image Segmentation

This GitHub repository contains the implementation of a real-time target detection system for Unmanned Aerial Vehicles (UAVs) using a finetuned version of YOLOv3, optimized network structure, feature extraction with Darknet53, and image segmentation techniques for payload delivery. The system achieves high precision, with an accuracy of 87% for general target detection and an additional 2% improvement specifically for small target detection.

Features

• Real-time target detection for UAVs
• Finetuned YOLOv3 model for high precision (87%)
• Enhanced small target detection for payload delivery (+2% accuracy)
• Image segmentation to identify regions with similar characteristics
• Optimized network structure and feature extraction using Darknet53

Performance

The target detection system achieves an impressive 87% precision on general target detection. For small target detection, it demonstrates a further 2% improvement, making it highly suitable for payload delivery tasks.

Dependencies

• Python 3.x
• PyTorch
• OpenCV
• Darknet53
• ArduPilot (for deployment on UAVs)