eyes-on-the-road Detect distracted drivers using a trained CNN. Some test images and their respective prediction labels (using ResNet-18) : Passing real-life images through the trained model : This is Mayukh, looks like he's texting. Training Metrics : The accuracy reaches 93.9% after 5th epoch. To Do : Record a video, pass the video frames through the model. Pass images through the model in real-time.