VEHICLE-MOUNTED REAL-TIME ANIMAL DETECTION SYSTEM FOR PREVENTING ROADKILL
by Jeesung Lee
Category: Social Science
Abstract – The paper compares SSD (Single Shot Multi-Box Detector)-Mobilenet v2 and YOLO (You Only Look Once)v4-tiny algorithms for their effectiveness in animal detection systems to address the growing roadkill issue. Current prevention methods, such as wild crossings, fences, and road signals, have limitations in cost or efficiency. Vehicle-embedded animal detection systems require rapid and resource-efficient algorithms. The study evaluated the algorithms on data from six different animals and found that YOLOv4-tiny outperformed SSD-Mobilenet v2 in real accident scene videos, making it a promising option for roadkill prevention.