Face detection method and system based on flexible convolution network

Abstract

The invention discloses a face detection method and system based on the flexible convolution network. The method comprises steps that an input image is received, and the image includes the to-be-detected face information; features of the image are extracted through the main network, and the features include semantic features and offset features; the semantic features and the offset features of theimage are respectively outputted through flexible convolution; according to the semantic features of the image, the face containing probability of an anchor box is outputted through the classification sub network; according to the semantic features of the image, the overlapping degree IOU between each prediction frame and a face is outputted through an IOU prediction branch; according to the offset features of the image, the face size and the position corresponding to each prediction frame are outputted through the position regression sub network. The method is advantaged in that the face inthe image can be accurately predicted.

Description

technical field [0001] The present invention relates to the interdisciplinary field of image processing and computer vision technology, and more specifically, to a face detection method and system based on a deformable convolutional network. Background technique [0002] Face detection is a popular research direction in the field of computer vision, and it is the basis of many face-related tasks, such as face recognition, face attribute recognition, face key point positioning, etc. Face detection is widely used in the field of computer vision, including video security, shopping mall monitoring, access control systems, and communication and entertainment. Traditional face detection mainly focuses on designing complex image features and training classifiers to achieve face detection. However, the method of manually designing features requires expert knowledge, and the detection effect needs to be improved. Deep learning abstracts the original image data layer by layer into t...

AI Technical Summary

Problems solved by technology

[0004] In view of the defects of the prior art, the purpose of the present invention is to solve the problem that the existing face detection based on convolutional neural network does not take into account the particularity of the features required for the position regression task, and the position regression requires features containing offset information, resulting in the The detection frame is not accurate, and the deformation modeling of the face is insufficient. Only outputting the probability of the face prediction frame will classify some backgrounds as faces, resulting in technical problems of false detection

Images