Akshat Sharma

During the first half of my tenure, i was responsible for developing Gaze Tracker using webcam by harnessing deep learning, Collected and preprocessed the data of 25 subjects and designed multiple models. After evaluting multiple models on the bases of accuracy and losses, the most optimized model showcased an accuracy of 92 percent. The model takes an input from the webcam and predicts the gaze coordinates on the screen. In the second half of my tenure, I perfected the drowsiness detection as I designed a CNN from scratch and collected data of 6 subjects and developed a highly accurate model with the accuracy of 98 percent (a perfectly fitted graph was achieved to prove this).