Eye Gaze Tracking Towards User Attention Analysis

Eye-tracking is a method that involves monitoring the position and movement of eyes that can help towards wide variety of applications including gaming, augmented reality, virtual reality, assistive technologies, human computer interface, marketing and medical diagnostic applications. The authors in this paper uses L2CS-Net, a novel gaze estimation architecture built on ResNet-50. The model, was validated on the MPII Gaze dataset through 5-fold cross-validation and exhibited consistent reduction in combined loss during training, showcasing its promising performance in accurate gaze estimation. The performance of the model was demonstrated for an application of profiling the attention of the user and monitoring the off-screen glances of the user while watching a video. Thus, demonstrating potential applications towards user attention detection in educational and marketing applications.