Method for generating human face from human eyes based on self-attention mechanism

Abstract

The invention discloses a method for generating a human face from human eyes based on a self-attention mechanism. The method for generating the human face from human eyes comprises the steps: extracting human eye information from a masked or shielded human face, fully mining an internal mapping relation between the human eyes and the human face through a self-attention mechanism adversarial generative network, synthesizing a vivid human face image according to the human eyes, and carrying out the face recognition of the generated human face. The method mainly comprises the following steps: constructing a data set for generating a human face from human eyes based on a human face image in a public human face data set, training the data set on a model provided by the invention, finishing training after adjusting parameters; extracting a human eye part in the face image and carrying out normalization processing; sending the information to a trained model to complete the generation of a human face; and finally, carrying out identity recognition verification on the synthesized human face and an original human face (ground truth). According to the method, the self-attention mechanism is introduced into the generative adversarial network to guide training, a more vivid face can be generated, the face recognition rate is effectively improved, and the method can be applied to the fields of public safety, terrorism and the like.

Description

technical field [0001] The invention designs a method for generating a human face from human eyes based on a self-attention mechanism, and relates to the technical fields of computer vision, deep learning and public security. Background technique [0002] With the advancement of face recognition technology, its application is becoming more and more extensive. At present, face recognition technology has a recognition rate of over 98% in some public face databases CelebA and LFW. However, in the actual application environment, the recognition effect of different scenes is also quite different, and the recognition effect is also very good in some places where the complete face can be clearly captured at close range, such as train stations, airports, examination rooms, mobile payment, etc.; In places where distance, light, background, occlusion and other factors interfere, the recognition effect is not satisfactory. For example, in the field of public security, criminals and sab...

AI Technical Summary

Problems solved by technology

However, in the actual application environment, the recognition effect of different scenes is also quite different, and the recognition effect is also very good in some places where the complete face can be clearly captured at close range, such as train stations, airports, examination rooms, mobile payment, etc.; In places where distance, light, background, occlusion and other factors interfere, the recognition effect is not satisfactory. For example, in the field of public security, criminals and saboteurs usually cover their faces, and only the information of their eyes can be seen, which is difficult for recognition. brings challenges

[0003] With the development of public monitoring, surveillance camer

Images