Attention-Guided Model for Robust Face Detection System

Face detection is a basic computer vision task which is required by various higher level applications including surveillance, authentication, and security system. To satisfy the demand on a high quality face detection method, this paper proposes a robust system based on deep learning model which utilize an attention-based training mechanism. This strategy enables the model to not only predicting the bounding boxes of faces but also outputs a heatmap that corresponds to the existence of faces on a given input image. The proposed method was trained on the most popular face detection dataset and the results show that it produces comparable performance to the existing state of the arts methods.