Deep Learning Enhancement of Road Safety: Detection of Distracted Drivers

This study attempts to identify driving instances using advanced techniques for deep learning. By using driver image analysis, we develop models which are capable of efficiently identifying and classifying distractions thus augmenting the general road safety.

Methodology:

1. Data collecting: compiling a dataset of driver images classified with several distraction levels
2. Data pre-processing is image data addition and normalisation meant to improve a model’s performance and resilience.
3. Development of Models: Image classification and distraction detection using CNNs
4. CNN model training uses the dataset to optimize hyperparameters for best
5. Evaluating a model using criteria including accuracy, precision, recall, F1

The results show that the CNN model in classifying different distractions effectively detects driving behaviors with great accuracy. This study intends to use knowledge to improve road safety and lower the number of accidents that are caused by driver distractions.