A face image synthesis method and device based on adversarial learning

Abstract

The invention relates to a face image synthesis method and device based on adversarial learning. The method provides a novel generative adversarial network (FC-GAN), and aims to synthesize a front face image under the standard illumination from an occluded face image. The FC-GAN uses a convolutional neural network structure an encoder-decoder as a generator, and two discriminators are introduced at the same time, one is a global discriminator used for distinguishing the authenticity of a whole face image, and meanwhile, the identity information of a person is kept unchanged; and the other oneis a local discriminator which is used for distinguishing the authenticity of the shielding area in the human face. In addition, a human face semantic segmentation network is introduced to strengthenthe consistency of five sense organs of a human face in the generated image. And by comparing the semantic tags of the synthesized face image and the real face image, the gradient is transmitted backto the generator to adjust the five sense organs of the synthesized face image. Experimental results on the reference data set Multi-PIE show that the FC-GAN performance is superior to that of most existing methods.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a face image synthesis method and device based on adversarial learning. Background technique [0002] Face recognition is an important research topic in the field of computer vision. Due to its ease of use, high accuracy, and non-intrusive features, it has shown great application potential in many fields such as security monitoring, financial services, mobile terminals, and physical businesses. In the past ten years, deep learning has achieved great success in face recognition, and the performance of face recognition systems based on deep neural networks has significantly exceeded that of face recognition systems based on manually designed features. Many start-up companies researching face recognition technology have sprung up like mushrooms after rain, such as Shangtang Technology, Megvii Technology, Cloudwalk Technology and Yitu Technology. [0003] The pe...

AI Technical Summary

Problems solved by technology

TP-GAN can only solve the problem of changing the perspective of the face, and cannot solve the problem of occlusion

Images