Browsing by Author "Muhammad Usman, Ahmad Yaqoob, Samiullah Qureshi and Abdul Hannan"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Comprehensive Evaluation of deep learning techniques for driver distraction detection using multi-source dataset
    (UMT, Lahore, 2024) Muhammad Usman, Ahmad Yaqoob, Samiullah Qureshi and Abdul Hannan
    This research introduces a novel approach to addressing the persistent issue of road accidents caused by driver distraction. The focus is on developing a real-time driver distraction detection system using artificial intelligence, specifically adapting the CNN architectures. The model incorporates an additional GAP2D layer, dense layers with ReLU activation, a dropout layer with a 40% dropout rate, and a final dense output layer with Softmax activation. Comprehensive training utilizes diverse datasets, including SF, DF, P’04, and BU aiming to establish a robust system for early detection of various driver distractions. The study’s objectives include minimizing distraction-related incidents across various settings, from driving scenarios to educational and workplace environments. The proposed system's range covers various industries, mainly aimed at safety, vigilance, and anti-accident. The significance of this study is based on the observation that road accidents are on the rise in all parts of the world and there is a focus on reducing the risk of deaths resulting from distracted driving. The paper is informative and has given a brief background of previous and existing techniques used in the identification of distraction. It is also listed in the Dataset description to indicate a range of driving situations and distractions that are encompassed by SF, DF, P’04 and BU. The rationale section of this paper explains why CNN architectures were selected for the model and presents the new changes and units introduced into the model. The proposed study also wisely describes what model training entails, for instance, data augmentation and validation set performance. Last but not least, the abstract comes with a list of references consisting of only the articles addressing the main concern of driver distraction detection.