Face Drawing Technology Based On Facial Features Research And Application

In recent years,relying on the innovation of artificial intelligence technology such as deep learning,face drawing technology has gradually become a research hotspot of experts and scholars,and is widely used in criminal investigation,digital media,education and training,film and television entertainment and other fields.The face drawing technology based on facial features obtains the correspondence between real face images and facial features through deep learning,and generates a two-dimensional image in real time according to the facial contour features drawn by the user,gradually approximating the face in the real world.However,the current face rendering technology relies on a large number of face feature data sets,and the generated face results have problems such as poor controllability and insufficient details.Therefore,this paper studies the face feature extraction method based on the improved Deeplab V3+ and A face image generation method based on an improved CGAN network,and an interactive face rendering system based on facial features is implemented.The main research results of this paper are as follows:(1)Aiming at the problem that the facial feature extraction method is difficult to obtain clear and stable facial features,this paper proposes a facial feature extraction method based on the improved Deep Lab V3+.Based on the network model of Deep Lab V3+,the network structure of the coding area is constructed to extract the features of the input image;the network structure of the decoding area is constructed to decode and output the semantic segmentation result;the ASPP module of the coding area is improved,and the depth-separable convolution is used instead of the ordinary convolution.And the BN layer is used to normalize the output,which reduces the training difficulty of the network.(2)In view of the problems of poor controllability of the generated results and insufficient details in the face generation method,a face image generation method based on an improved CGAN network is proposed.The generator and discriminator are constructed based on the CGAN network model,the generator structure is improved,and the detail generator is expanded on the UNet structure,which captures the color features of the input data.Add a spatial self-attention mechanism to the generator and discriminator to capture the global structured features of the input data.Design the loss function,train and output the model,and achieve high-realistic face generation.(3)Aiming at the problems of high editing threshold,interactivity and poor rendering effect in the current face rendering system,this paper designs and implements an interactive face rendering system based on facial features.According to the functional requirements of the system,it is divided into a face feature extraction module,a face image generation module,an interactive editing module and a basic component module.It has been verified that the system has a good interactive experience,a low editing threshold,and a strong sense of realism in the drawing results,which can meet the needs of practical applications.