Research On Face Sketch Generation Based On Generative Adversarial Network

As a commonly used method of criminal investigation,simulated portrait technology plays a huge role in public security work.However,the traditional hand-painted simulated portrait technology faces difficulties such as high requirements for painting skills and difficult modification.Therefore,it is difficult to popularize and apply this technology on a large scale.Although the computer face combination simulation portrait technology through the facial features database can effectively improve the drawing efficiency,the lack of diversity and intuition of the simulated portrait after combination restricts the further development of this technology.The powerful image processing capability of deep learning provides new methods and new ideas for simulating portraits,but it is still limited by the lack of face photos-hand-drawn sketch image datasets and cannot develop rapidly.Therefore,based on generative adversarial networks,this paper proposes a face sketch generation method based on face sketch image generation technology and sketch simplification technology.Specifically:First of all,aiming at the face sketch generation technology,this paper proposes a face sketch image generation algorithm based on Pix2Pix network structure with the help of generative confrontation network model,and improves it in terms of overall framework,network structure,loss function and training method,so that it can better cope with the task of face sketch generation and generate high-quality face sketch images.It is verified by experiments that this method works well.Secondly,for the sketch simplification task,this paper designs a generative adversarial network model dedicated to the sketch simplification task with the help of the generative adversarial network and the residual network,and according to the characteristics of the sketch image and the characteristics of the convolutional neural network,a fusion A hybrid attention mechanism of spatial attention mechanism and channel attention mechanism is proposed and applied to the sketch simplification task.It has been verified by experiments that both the face age estimation network and the optimized face age editing model have achieved good results.Finally,based on the face sketch generation technology and sketch simplification technology,this paper designs a method that can generate face sketches.By connecting the two technologies in series,the method can generate a face sketch with a simple enough picture,a relatively complete structure and a hand-painted style based on the input face photo.It is verified by experiments that the face sketch generated by this method has achieved good results both in appearance and in actual use.