Generative Adversarial Networks for Image Generation

ISBN-13: 9789813360471 / Angielski / Twarda / 2021 / 77 str.

Xudong Mao; Qing Li

Spis treści

Generative Adversarial Networks (GANs)

o   Introduction to GANs

o   Challenges of GANs

•    GANs for Image Generation

o   Image Generation

§  Overview of Image Generation

§  Single Domain Image Generation

§  Multi-domain Image Generation

o   Improving quality for Generated Image with LSGANs

o   Improving Training Stability: Theoretical Analysis

o   Multi-domain Image Generation with RCGANs

•    More Key Applications of GANs

o   Image to Image Translation

o   Unsupervised Domain Adaptation

o   GANs for Security

•    Discussion and Conclusion

•    Bibliography

O autorze

Xudong Mao is currently a Postdoctoral Fellow at the Hong Kong Polytechnic University. His research interests are in the areas of computer vision and deep learning, especially generative adversarial networks and unsupervised learning. His research work has been published in top-ranked journals and conferences in the area, such as TPAMI, ICCV, and IJCAI. Dr. Mao’s paper ‘Least squares generative adversarial networks’ has, to date (November 2020), been cited more than 1700 times since it was published in 2017 at the ICCV conference.

Qing Li is currently a Chair Professor at the Hong Kong Polytechnic University. He also serves/served as a Guest Professor of Zhejiang University, an Adjunct Professor of the University of Science and Technology of China, and a Visiting Professor at the Wuhan University and the Hunan University. His research interests include database modeling, multimedia retrieval and management, social media computing and e-learning systems. Dr. Li has published over 400 papers in technical journals and international conferences in these areas, and is actively involved in the research community by serving as a journal reviewer, program committee chair/co-chair, and as an organizer/co-organizer of numerous international conferences. Currently he is the Chairman of the Hong Kong Web Society, a councillor of the Database Society of Chinese Computer Federation (CCF), a member of the CCF Big Data Experts Committee, and a member of the international WISE Society’s steering committee.

Blurb

Generative adversarial networks (GANs) were introduced by Ian Goodfellow and his co-authors including Yoshua Bengio in 2014, and were to referred by Yann Lecun (Facebook’s AI research director) as “the most interesting idea in the last 10 years in ML.” GANs’ potential is huge, because they can learn to mimic any distribution of data, which means they can be taught to create worlds similar to our own in any domain: images, music, speech, prose. They are robot artists in a sense, and their output is remarkable – poignant even. In 2018, Christie’s sold a portrait that had been generated by a GAN for $432,000.

Although image generation has been challenging, GAN image generation has proved to be very successful and impressive. However, there are two remaining challenges for GAN image generation: the quality of the generated image and the training stability. This book first provides an overview of GANs, and then discusses the task of image generation and the details of GAN image generation. It also investigates a number of approaches to address the two remaining challenges for GAN image generation. Additionally, it explores three promising applications of GANs, including image-to-image translation, unsupervised domain adaptation and GANs for security. This book appeals to students and researchers who are interested in GANs, image generation and general machine learning and computer vision.