Image Analysis System to Identify Ineligible Content: Development and Software Implementation

This article presents the development and software implementation of image analysis system to identify ineligible content. The system leverages the YOLO v8 (You Only Look Once) object detection architecture and a dataset of 1630 images representing commonly associated ineligible symbols in the Russian Federation, particularly relevant to the Republic of Crimea. The system is designed to identify and classify these symbols with high accuracy, using a trained neural network model. The authors developed a comprehensive model for analyzing objects in images and videos, incorporating metrics such as object area, focus center, number of objects, confidence levels, and variance. These metrics provide insights into the contextual significance of detected objects and assist in identifying potentially illegal content. The software implementation of the system is detailed, including its structure, modules, and functionality for analyzing individual images and sets of images. Test scenarios demonstrate the system’s ability to accurately detect and analyze ineligible content. The system holds significant potential for monitoring and combating the spread of ineligible materials online. The work was carried out at the Center for Artificial Intelligence and Big Data Analysis of the V.I. Venadsky Crimean Federal University under the supervision of PhD in Information Technology Marina Rudenko.