Застосування глибоких нейронних мереж для розпізнавання обличчя водія у відеопотоці для забезпечення безпеки водія

Abstract

The paper considers the application of deep learning methods, in particular a combination of convolutional and recurrent neural networks, for driver face recognition in a video stream. The main focus is on developing an approach that provides high accuracy and recognition speed for integration into vehicle safety systems. The proposed method uses CNNs to process individual video frames for facial feature extraction and RNNs to analyze temporal dependencies, such as blinking and eye closure duration, as well as to predict gaze direction.