Driver Behavioral Detection & Alert System
[1]Khaja Moinudeen AK, [2]Kruthik J, [3]Madhu Sudhan S,[4]Mohammed Saif,
[5]Mr. Shreenidhi B S
[1][2][3][4] UG Students, DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING
[5] Asst Professor, DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING, Dayananda Sagar Academy of Technology and Management, Bangalore, India
Abstract
Driving involves a collection of actions that demand intense focus. Sometimes other actions like eating, drinking, talking, making phone calls, adjusting the radio, or being sleepy take precedence over these behaviors. Additionally, these are the primary reasons for today's traffic accidents. Therefore, creating application that alerts the driver beforehand is crucial. In this study, a lightweight convolutional neural network architecture along with Haar-cascading and Facial landmark is introduced to identify driver behaviors, assisting the warning system in correct information delivery and traffic collision minimization. Combining feature extraction and classifier modules creates this network. The feature extraction module extracts the feature maps by utilising the benefits of average pooling layers, depthwise separable convolution layers, standard convolution layers, and proposed adaptive connections.
The feature extraction module, which directs the network in learning the salient features, makes use of the benefit of the convolution block attention module. To determine the probability of each class, the classifier module uses a global average pooling and softmax layer. The whole architecture keeps classification accuracy while optimising the network parameters. The Module is trained using YOLO, to train a YOLO model, you first need to prepare a dataset of labeled images. The dataset should include images of the objects you want the YOLO model to detect, along with the corresponding bounding box annotations and class labels. once the YOLO model is trained, it can be used to quickly and accurately detect objects
