Аналіз вибору героїв DOTA 2 в реальному часі по зображенню вже обраних героїв

Abstract

This paper investigates the development of a real-time hero draft evaluation system in Dota 2 based on computer vision and mathematical modeling methods. The proposed approach focuses on automatic recognition of selected heroes from graphical user interface images during the drafting phase. The system analyzes and interprets visual data extracted from the game screen and transforms it into structured representations suitable for further strategic evaluation. The work examines practical applications of convolutional neural networks and real-time object detection algorithms for identifying heroes in draft screenshots or video streams. Special attention is given to the integration of computer vision techniques with statistical modeling and elements of game theory in order to assess team synergy, counter-pick efficiency, and role balance.