Detekce osob na dopravním prostředku pomocí YOLO algoritmu

Abstract

The bachelor thesis deals with the problem of detection of people on the vehicle using the YOLO algorithm. In the introduction, the basics of machine and deep learning, types of machine recognition, types of detection algorithms, history, principle and function of YOLO algorithm including comparison of different versions are explained. Based on the information, the necessary datasets and programs are developed to learn a custom detection model for detecting people on a vehicle. The learned detection model is verified on a test video file. Based on the verification, the created dataset is then augmented with collages and other detection models are trained and compared with each other. The actual work concludes with an evaluation of the detection models and a description of the sequence of events to enable the training of a custom detection model.