Systém pro detekci rukou a gest

Abstract

This master’s thesis deals with the development of a custom hand and gesture detection system. The first part of the thesis describes machine learning methods, neural networks and convolutional neural networks. Subsequently, an analysis of publicly available gesture detection libraries was conducted. Based on the optimal results, two classification models were developed, capable of recognizing five classes of gestures in real-time. The system also implements a TCP/IP communication interface for data distribution to the external applications in JSON format, which was experimentally tested on a real system.