Neural Network Based Scope Positioning System for First-Person Shooters

Abstract

In this paper, we present a new approach to assessing the position of the sight in first-person shooters. In such games, one of the most difficult tasks is to accurately aim the weapon. And if the game has a bullet drop, you have to take into account the properties of the bullet and weapon and shoot higher. Position of the scope depends also on the distance to the target and angle. Traditional methods of software-assisted aiming require access to the game files or processes. In our approach, we use only screen captured images.We present our system of automatic aiming based on neural networks. The main advantage of this approach is that our system for accurate definition of scope position requires only information about what is happening on the screen. To achieve this goal, we use technologies of convolutional neural networks that allow detecting objects in the image, as well we used Feed-Forward neural networks for adjusting the sight.