Attention monitoring using eye gaze with a hybrid ensemble learning approach

Healthcare, education, transportation safety, and human-computer interaction are just few of the many tasks that require monitoring. Monitoring is important for all of these tasks and more. This paper presents innovative work that has been done in the field of attention monitoring. The work involves combining a hybrid eye gaze model with deep learning in order to monitor the level of attention that a driver is paying. This paper provides a description of the hybrid eye gazing model that was suggested, as well as the results that were produced by the model. Data augmentation techniques such as rotation, shifting, shearing, and flipping are used to the proposed model, along with adjustments such as changing the fill mode in terms of zooming into the image and rescaling. The suggested model makes use of an augmented dataset. In order to ensure that the model is trained in a reliable and consistent manner, all of these aspects are essential. Modern pre-trained architectures, such as VGG16, VGG19, InceptionV3, EfficientNetB0, EfficientNetB7, and InceptionResNetV2, are the foundation upon which our model is constructed. These designs are modified, and then more layers are added, in order to facilitate the process of capturing very minute attention dynamics using them. The accuracy and effectiveness of the model were later improved by the utilization of a model ensemble. After some time has passed, the XGBoost model is amalgamated with all of the other models that were utilized previously in the hybrid model technique. This is done in order to achieve improved accuracy and efficiency of the model. Many different assessment measures, such as accuracy, precision, recall, F1 Score, and support, are utilized in order to conduct an adequate evaluation of the performance of the model. Through the use of these indicators, a comprehensive comprehension of the model's capacity to identify and forecast attention patterns in a variety of locations is achieved. We were able to obtain the highest level of accuracy from VGG19 and InceptionResNetV2 after utilizing the models, which was 84.6% and 83.6% respectively. A score of 82% was achieved by the VGG16 hybrid models during the accuracy test. The Hybrid Eye Gaze Model is a powerful and adaptable attention monitoring system that can be utilized for a wide range of applications. It utilizes deep learning and pre-trained architectures.