摘 要：目前，学者们对人物图像生成技术的研究主要集中在对人物姿势的编辑方面，忽略了身体的外观特征，导致所生成人物图像的质量不够理想。鉴于此，提出一种融合 CoT Block 的人物图像生成方法，即在已有 PG2 模型的基础上，将改进后的 CoT Block 引入到生成对抗网络中，通过对上下文语义信息的挖掘以及结合自注意力学习机制，更好地捕获人体姿态特征；然后利用 PGGAN 中的鉴别器进一步增强对图像真伪的鉴别能力。实验结果表明，改进后的算法有效提高了人物图像的生成质量。
关键词：图像生成；生成网络；CoT Block；PGGAN
DOI:10.19850/j.cnki.2096-4706.2023.07.023
中图分类号：TP391 文献标识码：A 文章编号：2096-4706（2023）07-0090-04
Character Image Generation Algorithm Fused with CoT Block
YAO Xingyue
(School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan 232001, China)
Abstract: At present, scholars' research on the generation technology of character image mainly focuses on the editing of character posture, ignoring the appearance characteristics of the body, resulting in the quality of the generated character image is not ideal. In view of this, a character image generation method fused with CoT Block is proposed, that is, based on the existing PG2 model, the improved CoT Block is introduced into the generative adversarial network, and the human posture features are better captured through the mining of context semantic information and the combination of self-attention learning mechanism; Then the discriminator in PGGAN is used to further enhance the ability to identify the authenticity of the image. Experimental results show that the improved algorithm can effectively improve the generation quality of character image.
Keywords: image generation; generative network; CoT Block; PGGAN
参考文献：
[1] GOODFELLOW I J，POUGET-ABADIE J，MIRZA M，et al. Generative Adversarial Nets [EB/OL].[2022-10-29].https://arxiv.org/ pdf/1406.2661v1.pdf.
[2] MA L Q，JIA X，SUN Q R，et al. Pose guided person image generation [J/OL].arXiv：1705.09368 [cs.CV].[2022-10-29].https:// arxiv.org/abs/1705.09368v3.
[3] ESSER P，SUTTER E. A Variational U-Net for Conditional Appearance and Shape Generation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City：IEEE， 2018：8857-8866.
[4] SIAROHIN A，SANGINETO E，LATHUILIERE S，et al. Deformable GANs for Pose- Based Human Image Generation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Salt Lake City：IEEE，2018：3408-3416.
[5] ZHU Z，HUANG T T，SHI B G，et al. Progressive pose attention transfer for person image generation [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. Long Beach：IEEE，2019：2342-2351.
[6] ZHANG J S，LI K，LAI Y K，et al. PISE： Person Image Synthesis and Editing with Decoupled GAN [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition（CVPR）. Nashville：IEEE，2021：7978-7986.
[7] LI Y H，YAO T，PAN Y W，et al. Contextual Transformer Networks for Visual Recognition [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2023，45（2）：1489-1500.
作者简介：姚星月（1998—），女，汉族，安徽蚌埠人，硕士在读，研究方向：图像处理。