Research on Face Detection Method Based on Deep Learning

Faster R-CNN is a popular method in object detection applications. Based on the similar characteristics of the category in face detection and the category label of target background in target detection, this paper improves the region suggestion network in Faster R-CNN method, and realizes multi-scale face detection by adding multiple detectors. We train and test the multi-scale face detection network model on the Wider Face dataset. Through the experimental results, it can be obtained that the accuracy of the multi-scale face detection method is 89.6%, which has practical application value.