A New Method for Detecting the Fatigue Using Automated Deep Learning Techniques for Medical Imaging Applications

Abstract

In the human race, Fatigue may contribute to a decline in efficiency. Fatigue is a risk factor for health and a component of quality degradation. The effects of Fatigue include sleep disorders, depression, and worry, all of which can lead to life-threatening issues. This project uses machine learning and deep learning techniques to identify a person's degree of Fatigue and its effects. The developed deep learning network can accurately distinguish between normal and exhausted states. This detection device uses physiological characteristics to guarantee high detection rates and accuracy. The project aims to recognize fatigue levels by examining the features extracted from the batch of images and classifying them into their respective class labels, such as Alert, Non-Vigilant, and Fatigued. To implement the detection procedures using deep structured learning that will yield very accurate image recognition and classification results, a large-scale image dataset is transported through efficient algorithm strategies and is processed to transform the data by labelling the patterns, tracking the correlations, and producing supreme results. The images will be diagnosed by employing the pre-trained models of the Convolutional Neural Networks (CNN), convolving through the hidden layers, applying the filters, and sharing the weights. Alex Net, Resnet50, and MobilenetV2 are the potential classifiers that will expand, filter, train, compress, and test through the neurons of the subjects. The layers and the non-linear functionalities are designed in the wake of the structured embedding of the model to deliver efficient metrics. The proposal offers the best accuracy for the established MobilenetV2 model with 99.8% accuracy and validates it with high-performance results.