ACCELERATING OBJECT DETECTION: DESIGNING A REAL-TIME YOLOV3 PROTOTYPE SYSTEM

Abstract

With the continuous development of knowledge and technology, artificial intelligence (AI) has emerged as a significant area of research. Machine Learning, a prominent approach in AI, attempts to mimic human behavior to solve problems and automate tasks. Central to machine learning is the training process, which requires data for learning. Object detection, a fundamental computer vision problem, involves recognizing objects in images or videos and is associated with numerous applications, including image classification. The You Only Look Once (YOLO) method has gained attention for its speed and accuracy in object detection, outperforming other algorithms by predicting bounded boxes and class probabilities directly from the full image.

This research aims to develop a real-time object detection system using the YOLO method. The process starts with the preprocessing of the original image, involving resizing, grayscale conversion, and edge detection convolution, ultimately leading to object recognition. The proposed system utilizes surrounding object data to enhance the algorithm's accuracy and performance.

To assess the effectiveness of the YOLO-based system, various experiments are conducted to measure accuracy and training speed. The study demonstrates that the YOLO method, while exceptionally fast, may exhibit localization errors and relatively slow training speeds. By incorporating surrounding object data, the research seeks to improve the algorithm's overall performance.