Detecting and alerting distracted drivers is essential to reducing traffic accidents and enhancing road safety. Although various deep learning-based methods have been proposed for distraction detection, most rely on single-perspective images, making them vulnerable to occlusion and environmental challenges. To overcome these limitations, this project proposes a novel driver distraction detection approach that integrates Multi-Task Learning (MTL) and Ensemble Learning. The system processes images captured from multiple driver perspectives and utilizes MTL to enable knowledge sharing between tasks. This improves the model's ability to generalize across different distraction types and driving scenarios. Detection results from each perspective are then combined through an ensemble strategy, enhancing the system’s overall accuracy and robustness. The proposed model is capable of identifying a wide range of distraction behaviors such as texting, phone use, and passenger interaction, with high reliability. Due to its ability to adapt to complex real-world conditions and provide consistent performance, the proposed method demonstrates significant potential for integration into driver monitoring and assistance systems, aiding the development of safer and more innovative transportation technologies.