Unmanned aerial vehicle (UAV) object detection using YOLO detector

• Main Author: Guo, Huanchen
• Other Authors: Yap Kim Hui
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/175475
• Institution: Nanyang Technological University

Nowadays, unmanned aerial vehicles (UAVs) have gained significant popularity i n computer vision (CV) and remote sensing (RS). Concurrently, GPU acceleration has facilitated the widespread adoption of deep l earning (DL). This dissertation comprehensively reviews mainstream object detection methods and relevant literature. YOLOv8, one of the SOTA (State Of The Art) detectors based on the deep learning framework, is implemented to fulfill the UAV based object detection task. Meanwhile, several representative one stage and two stage detectors are also tested for comparison . The effectiveness of setting different input image sizes and batch sizes for training is also evaluated and discussed. Inspired by experiments conducted in this dissertation and related literature, the baseline of YOLOv8 is modified and the modified version achieves 2% ~ 5% higher accuracy.