An Intelligent System for Driving Distraction Identification by Fine-Tuned Modified Transfer Learning Techniques

Driving faults resulting from unmatured driving behaviors, disobedience of customs and laws, and a deficiency of driving expertise remain a serious global public safety risk. To solve these problems, the national authorities must correctly implement the legislation and give accurate driver warnings that need remedial action. This study explores the complex relationship between human behavior and the various elements that contribute to roadside diversions. A robust artificial intelligence system based on transfer learning is proposed to tackle these problems. The work involved the creation of a new data set containing ten distinct driving habits, which were subsequently classified using enhanced and modified Densenet201 and VGG19 models. An extensive analysis was carried out to determine how well-performing popular matrices performed on the updated VGG19 and DenseNet201 models in terms of the confusion matrix, f1 score, accuracy, precision, and recall. As seen by its 99.90% recall, F1 score, accuracy, and precision, the improved DenseNet201 model performed exceptionally well after integrating the Adam optimizer. These results show that the recommended approach is successful in detecting instances of driving while inattentive, resulting in a notable improvement in detection accuracy.