科研成果详情

题名GC-GAN: Photo Cartoonization Using Guided Cartoon Generative Adversarial Network
作者
发表日期2023
会议名称32nd International Conference on Artificial Neural Networks
会议录名称Artificial Neural Networks and Machine Learning – ICANN 2023: 32nd International Conference on Artificial Neural Networks, Heraklion, Crete, Greece, September 26–29, 2023, Proceedings, Part V
会议录编者Lazaros Iliadis, Antonios Papaleonidas, Plamen Angelov, Chrisina Jayne
ISBN978-3-031-44191-2
ISSN0302-9743
卷号Lecture Notes in Computer Science (LNCS, volume 14258)
页码65-77
会议日期September 26–29, 2023
会议地点Heraklion, Greece
出版地Cham
出版者Springer
摘要

Image cartoonization, a subfield of image translation, has gained increasing attention in recent years. However, the overall performance of existing methods is limited by shortcomings such as poorly fitting regions. To address these issues, our paper proposes the Guided Cartoon Generative Adversarial Network (GC-GAN). Our approach introduces a segmentation step before the training process, which splits and guides mixed training images into a human face set and a scenery set. This enables our model to extract features specifically from cartoon faces and generate more realistic results. Furthermore, we include a loss function called triplet loss in our framework, which drives the network to bring output closer to a referenced image and focus more on the detailed parts of the training images. This improves the overall quality of the generated images and addresses the issue of poorly fitting regions. Compared to the state-of-the-art White-box CartoonGAN, our work improves Fréchet Inception Distance by 18.7% and Kernel Inception Distance by 17.6%, respectively. Additionally, our work surpasses AnimeGAN in terms of Fréchet Inception Distance by 40.7% and Kernel Inception Distance by 47.4%.

关键词Generative adversarial network Image cartoonization Image translation
DOI10.1007/978-3-031-44192-9_6
URL查看来源
收录类别CPCI-S
语种英语English
WOS研究方向Computer Science
WOS类目Computer Science, Artificial Intelligence ; Computer Science, Theory & Methods
WOS记录号WOS:001156948900006
Scopus入藏号2-s2.0-85174633984
引用统计
文献类型会议论文
条目标识符https://repository.uic.edu.cn/handle/39GCC9TT/11673
专题理工科技学院
通讯作者Chen, Donglong
作者单位
1.Guangdong Provincial Key Laboratory IRADS,BNU-HKBU United International College,Zhuhai,China
2.Fudan University,Shanghai,China
第一作者单位北师香港浸会大学
通讯作者单位北师香港浸会大学
推荐引用方式
GB/T 7714
Zhang, Jiachen,Hou, Huantong,Chen, Jingjinget al. GC-GAN: Photo Cartoonization Using Guided Cartoon Generative Adversarial Network[C]//Lazaros Iliadis, Antonios Papaleonidas, Plamen Angelov, Chrisina Jayne. Cham: Springer, 2023: 65-77.
条目包含的文件
条目无相关文件。
个性服务
查看访问统计
谷歌学术
谷歌学术中相似的文章
[Zhang, Jiachen]的文章
[Hou, Huantong]的文章
[Chen, Jingjing]的文章
百度学术
百度学术中相似的文章
[Zhang, Jiachen]的文章
[Hou, Huantong]的文章
[Chen, Jingjing]的文章
必应学术
必应学术中相似的文章
[Zhang, Jiachen]的文章
[Hou, Huantong]的文章
[Chen, Jingjing]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。