Abstract
Face detection in unconstrained environments is a challenging problem due to partial occlusions with pose variations. Existing partial occluded face detection methods require training several models, computing hand-crafted features, or both. In this paper, our contributions are two-fold. First, we propose our Large-Scale Deep Learning (LSDL), a method that requires a single Convolutional Neural Network (CNN) model without computing any hand-crafted features to detect faces. The model is trained with a large number of face training examples that cover most partial occlusions and non-partial occlusions facial appearances to detect unconstrained multi-view partially occluded and non-partially occluded faces. The LSDL face detection method is achieved by selecting detection windows with the highest confidence scores using a threshold. Second, we introduce new four image datasets consisting of large-scale labeled face dataset, noisy large-scale labeled non-face dataset, CrowdFaces dataset, and CrowdNonFaces dataset intended to be used for face detection training. Our evaluation results show that LSDL achieves the best performance on AFW dataset and a comparable performance on FDDB dataset compared to state-of-the-art face detection methods without manually extending or adjusting the square detection bounding boxes.