A simulation portrait generation method and device based on deep learning

Abstract

The invention discloses a simulation portrait generation method based on deep learning. The simulation portrait generation method comprises the following steps of S1, obtaining a face splicing pictureor a face contour picture; S2, generating a face image, and respectively using the face splicing image or the face contour image as condition input of a conditional generative adversarial network forconstraining the network to generate more realistic and more natural complete face images which are respectively similar to the face splicing image and the face contour image; and S3, carrying out the face identification and editing, identifying the generated complete face image by the victim, putting forward local or overall modification suggestions, editing or regenerating the image according to the modification suggestions, and finally outputting the face image. The invention further discloses a device corresponding to the method. According to the simulated portrait generation method provided by the invention, a conditional generative adversarial network algorithm and a convolutional neural network algorithm are introduced based on improvement of an existing hand drawing method and a splicing type simulated portrait method, so that the generated simulated portrait is more vivid, natural, lifelike and interesting, and the practical effect can be greatly improved.

Description

technical field [0001] The invention belongs to the technical field of intelligent criminal investigation, and more specifically relates to a method and device for generating simulated portraits based on deep learning. Background technique [0002] With the wide coverage of the police case-handling process by various media and the general improvement of people's education level, criminal suspects study the police investigation and case-solving process through various information channels, and their anti-investigation capabilities are gradually improving. Even today with a wide-ranging monitoring network and highly accurate fingerprint and DNA identification technologies, criminal suspects can still destroy clues and evidence by avoiding monitoring, destroying monitoring facilities, erasing fingerprints, and cleaning crime scenes. , It has caused great difficulties for the police to investigate and solve the case. In such cases, the deep impression of the criminal suspect's ...

AI Technical Summary

Problems solved by technology

However, this automatic analog portrait also has some shortcomings and deficiencies. For example, the spliced ​​portraits often appear blunt and not vivid enough, and it is difficult to guarantee the effect of use.

Images