Synthetic Face Images with ST-ED from MPIIFaceGaze and EYEDIAP

This is a dataset with synthetic full-face images generated from MPIIFaceGaze and EYEDIAP (Continuous Screen Target) datasets with the ST-ED method. These full-face images are generated with desired gaze directions for both eyes and the desired head pose. All head poses in this dataset are inherited from the normalized face images (with rolls of the head pose removed) of MPIIFaceGaze and EYEDIAP (Continuous Screen Target). The Multi-PIE 68-points markup facial landmarks are estimated based on the PRNet for every full-face image included in this dataset.