Detekce zdravotních problémů řidiče

Abstract

The main focus of proposed master thesis is to explore a different ways of driver's health problems detection by using machine learning. The experiments are dedicated to detection from image data, derived body and facial keypoint features and mixture of both mentioned ways are presented. The experiments follows each other with the goal of outperform the previous ones. Various metrics (F1 and Kappa score, accuracy and processing time) are provided to allow experiment comparison. The attention is also on tuning machine learning models, dataset processing and practical implementation.