Browsing by Author "Ayesha Batool, Hammad Hassan and Hamza Dilshad"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    REAL-TIME DRIVER DISTRACTION DETECTION APPLICATION
    (UMT, Lahore, 2024) Ayesha Batool, Hammad Hassan and Hamza Dilshad
    Thus, this research uses the reliable CNN architectures to train a complex model for enhancing driver distraction detection. The datasets we have used are State Farm Distracted Drivers, DrivFace, pointing'04 and Boston University, which involve numerous scenes for a general evaluation of our approach. As we already know, CNN architectures are one of the best models for image classification, and it forms the basis for feature extraction and learning. The pre-training on these architectures is done on the Image Netdataset, which helps incorporate a wide range of low-level and high-level features for transfer learning. Additional specification of distraction scenarios in terms of the specific characteristics narrows the model's discriminator. Careful hyper-parameters optimization checks for convergence to minimize the risk of overfitting, thus improving the model outcomes. This work has been applied in Python using Tensor Flow and Keras, and its performance was assessed depending on parameters such as accuracy, precision, recall, and F1. The current paper offers a detailed account of distracted driver identification in actual conditions that demonstrates how these architectures with ImageNet weights can effectively detect distracted drivers