Project focuses on detecting distracted driving behaviors using machine learning models. By leveraging MobileNetV2 and EfficientNet-B0, we classify driver distractions, such as texting or talking on the phone, from 2D dashboard camera images. This approach aims to improve road safety and reduce distracted driving risks.
The project utilizes the State Farm Distracted Driver Detection Dataset, which captures real-world driver behaviors in a controlled setting. It includes images labeled across 10 predefined classes of distracting activities, enabling effective model training for driver behavior classification.
The dataset is well-balanced, with approximately equal representation of each class, and includes images captured under varying lighting conditions to simulate real-world scenarios. Its moderate size makes it ideal for transfer learning, leveraging pre-trained models and fine-tuning them for this specific task.
This methodology enhances model robustness, allowing accurate classification of driver behaviors even with limited data.