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.