Key‐point‐guided adaptive convolution and instance normalization for continuous transitive face reenactment of any person

• Published in: Computer animation and virtual worlds Vol. 35; no. 3
• Main Authors: Xu, Shibiao, Hua, Miao, Zhang, Jiguang, Zhang, Zhaohui, Zhang, Xiaopeng
• Format: Journal Article
• Language: English
• Published: Chichester Wiley Subscription Services, Inc 01.05.2024
• Subjects: Convolution; face reenactment; human‐centered computing; Image contrast; Image reconstruction; visualization; visualization application domains
• ISSN: 1546-4261; 1546-427X
• DOI: 10.1002/cav.2256

Face reenactment technology is widely applied in various applications. However, the reconstruction effects of existing methods are often not quite realistic enough. Thus, this paper proposes a progressive face reenactment method. First, to make full use of the key information, we propose adaptive convolution and instance normalization to encode the key information into all learnable parameters in the network, including the weights of the convolution kernels and the means and variances in the normalization layer. Second, we present continuous transitive facial expression generation according to all the weights of the network generated by the key points, resulting in the continuous change of the image generated by the network. Third, in contrast to classical convolution, we apply the combination of depth‐ and point‐wise convolutions, which can greatly reduce the number of weights and improve the efficiency of training. Finally, we extend the proposed face reenactment method to the face editing application. Comprehensive experiments demonstrate the effectiveness of the proposed method, which can generate a clearer and more realistic face from any person and is more generic and applicable than other methods. This work presents a continuous transitive face reenactment algorithm that uses face key points information to gradually reenact faces based on two stages GAN, which contains the key face points transformation module and the facial expression generation module. The process involves transforming key points from the source face and generating corresponding facial expressions on the target face.