Eye Gaze Correction System

The process of gaze correction involves manipulating the eye region of an individual in order to redirect their gaze towards the camera. This method, which is a particular kind of image resynthesis, may be used to take pictures where the subject appears to be looking straight at the camera, among many other practical uses. Using an inpainting model, we present a novel method for accomplishing gaze correction in this research. With our method, we train a face-to-face learning model on photos, then manually fill in the blanks where the eyes should be. A benefit of our approach is that it does not necessitate labelled training data that provides information about the exact location of the head and the orientation of the eyes. As a result, we had a rather easy time amassing the training data we needed. To retain the eye area identification information in the original input, we present a self-guided pretrained model that can learn angle-invariant properties. Our tests show that, when applied to a dataset gathered from multiple websites, our approach produces incredibly convincing gaze-corrected results. In our study, we go into detail about this dataset.