Face Detection Method Based on Improved Convolutional Neural Network

Nowadays, the number of Internet image data is increasing rapidly, so it is urgent to understand image information by corresponding intelligent system. The purpose of face detection is to determine the location and size of the face in an image, which is one of the main research directions in the field of computer vision. Under certain constraints, face detection technology has achieved very high performance, but under the unconstrained conditions, many existing algorithms just change the network model structure, but can't meet the actual needs. In view of the above problems, this paper proposes a face detection method region based on multi-resolution full convolution neural network and convolution neural network. In this method, from the multi-scale point of view, a multi-resolution sliding window is used to generate a multi-level resolution human face heat map. According to the local hottest region on the heat map, the face candidate region is obtained. Finally, the face candidate region is sent to CNN classification network for classification, and the face position is obtained. The experimental results show that, the method proposed in this paper improves the detection rate of human face effectively and has better practicability.