Driver Behavioural Detection & Alert System
Khaja Moinudeen AK, Kruthik J, Madhu Sudhan S,Mohammed Saif, Mr. Shreenidhi BS
Student, DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING, Dayananda Sagar Academy of Technology and Management, Bangalore, India
Abstract
Driving requires a series of actions that call for significant concentration. Sometimes these behaviours are put on the back burner in favour of other activities including smoking, eating, drinking, talking, making phone calls, adjusting the radio, or falling asleep. These are also the main causes of today's traffic accidents. As a result, developing apps
It is essential to notify drivers in advance. To help the warning system convey accurate information and reduce traffic collisions, a lightweight convolutional neural network architecture is developed in this study.
This network is produced by combining feature extraction and classifier modules.
By utilising the advantages of average pooling layers, depthwise separable convolution layers, standard convolution layers, and suggested adaptive connections, the feature extraction module extracts the feature maps.
The benefit of the convolution block attention module is utilised by the feature extraction module, which guides the network in learning the important features. The classifier module makes use of a global average pooling and softmax layer to calculate the probability of each class. Throughout the architecture, classification accuracy is kept while network parameters are optimised.
The entire network is trained and tested using three benchmark datasets: the State Farm Distracted Driver Detection, the American University in Cairo version 1, and the American University in Cairo version 2. Since there are ten classes, the overall class accuracy is 99.95%, 95.57%, and 99.61%, respectively. A number of video tests were also conducted in HD (High Definition), FHD (Full High Definition), and VGA (Video Graphics Array) resolutions; you can view them at https://bit.ly/3GY2iJl.
