Real-Time Detection of Driver Distraction Using Hybrid Model VGG16 and Random Forest A Deep Learning Approach

Distracted driving causes a large number of traffic incidents worldwide. Mobile devices, in-vehicle technologies, and other external stimuli have made it harder for drivers to focus on the road. This study seeks an accurate driver preoccupation detection system. We want to construct an automated drowsiness diagnosis system by categorizing facial expression photographs as “Open Eyes,” “Closed Eyes,” “Yawn,” and “No Yawn.” These facial states are accurately classified using convolutional neural networks. New data augmentation on 2899 images improved our model's generalizability and durability. This paper employs eye fixation data, a Convolutional Neural Network for Temporal Analysis, and a Random Forest classifier to build a hybrid model that uses VGG16 features to improve distraction detection. We employ a CNN prediction model to analyze and extract dataset features in real-time. The CNN model accurately detects driver tiredness by distinguishing open and closed eye states. The average accuracy is 92%. A high training accuracy means that the model learns properties well, while a good testing accuracy means that it can apply to new data.