Fig. 1
From: An investigation of privacy preservation in deep learning-based eye-tracking
Eye gaze: illustration of the main part of the target model, which is the focus of the attack: FCs refer to different fully connected layers, while CNNs are convolutional neural network parts. After face and eye detection with regression tree, the left-eye and the right-eye are fed into CNN-E, which is CNN for eyes (shared weights) and a separate CNN, where face crop is the input (CNN-F). The photograph of the face is a modified from a publicly available image [28] under the Unsplash License [29]