Double-branch Anchor Free face detection method and system based on convolutional neural network

Abstract

The invention discloses a double-branch Anchor Free face detection method and system based on a convolutional neural network. The method comprises the following steps: establishing an Anchor Free face detection neural network of a double-branch structure; preprocessing an open source face detection training set Wider Face to obtain a training set; training the neural network by using the data set to obtain a trained face detection neural network model; inputting a to-be-detected picture into the trained model for processing to obtain a preliminary detection result; and filtering a repeated detection frame in the detection result by using a scale selection module. According to the invention, face detection is carried out only by using a full-convolution neural network without complex anchor frame design and post-processing steps, so that the post-processing steps of a face detection algorithm are simplified and the detection speed is improved on the premise of ensuring the detection accuracy.

Description

technical field [0001] The invention belongs to the technical field of computer digital image processing and pattern recognition, relates to the field of face detection, in particular to a double-branch Anchor Free face detection method and system based on a convolutional neural network. Background technique [0002] Face detection technology refers to the automatic capture of human faces in videos or images by computer, which is a key technology in the field of modern computer vision. In recent years, with the rapid development of artificial intelligence technology and high-performance computing equipment, face detection has been widely used as a frontier step in tasks such as automatic driving, liveness detection, expression analysis, and identity authentication. However, in unconstrained scenes, there are huge differences in scale, pose, and skin color between different faces, which brings great challenges to detection. [0003] Face detection is a classic task in comput...

AI Technical Summary

Problems solved by technology

However, there are some natural defects in the anchor box mechanism, including: (1) the detection performance is strongly related to the design of the anchor box, and different anchor boxes need to be carefully designed for different business scenarios, which is not conducive to the rapid migration of the algorithm; (2) the anchor box The frame mechanism relies on the NMS algorithm to filter redundant detection frames, but the NMS algorithm is difficult to parallelize, which seriously affects the detection speed and is not conducive to practical applications

Images