DRIVER DRIVING SCORE CALCULATION AND DROWSINESS DETECTION USING DEEP LEARNING

Abstract

This initiative was created in order to prevent accidents caused by drowsiness. When the drowsiness in a driver is detected, the system notifies them right away. Subsequently, it records the date, time, and duration of the drowsiness in a database to create a score that indicates how well, the motorist is driving. The project aggressively approaches drivers with high scores to reduce the likelihood of future accidents because prevention is more effective than therapy. To identify drowsiness in real- time, a custom CNN model is used in this project. The Model achieve 98.387% accuracy on the testing dataset. To achieve optimal accuracy under all conditions, the training dataset contains low-light, hazy, and spec as well as reflection-filled images and this is done by using the Media Research Lab(MRL) eye database. It allows the model to detect drowsiness in real-time video with maximum accuracy under all circumstances.

Keywords: convolutional neural network drowsiness, hyperparameters MySQL