Independent Artist Dagu Spitzer Explores PaddleGAN

Once praised by the “father of convolutional networks” Yann LeCun as “one of the most interesting ideas in computer science over the past decade”, GANs – Generative Adversarial Networks – have become one of the most popular deep learning algorithms in recent years, capturing the attention of AI researchers worldwide.

The popularity of GANs stems from their ability to showcase the magical process of creating something out of nothing, shining in various application domains and giving rise to many fancy applications.

Applications such as image generation/repair, super-resolution, image noise reduction, outfit/makeup changes, image style transfer, text/audio generation, etc., cover the internet, entertainment, gaming, and industry, bringing unprecedented joy to the dull research life and adding color to mundane life.

Many of the popular applications you may be familiar with rely on the powerful capabilities of GANs.

Video/Photo Restoration

Using a series of GAN capabilities like super-resolution, colorization, and frame interpolation, you can travel back to 100 years ago in Beijing without learning magic, experiencing the historical atmosphere and cultural life of old Beijing.

Check out the complete GitHub tutorial and principles:

https://github.com/PaddlePaddle/PaddleGAN/blob/master/docs/zh_CN/tutorials/video_restore.md

Super Resolution

GANs can also be used to enhance the resolution of images and videos, which can not only restore old audiovisual materials but also allow developers to transmit low-resolution audiovisual content during video calls/transmissions to reduce bandwidth requirements and improve transmission speed, cleverly restoring the content at the terminal using super-resolution.

AI Face Swap

Like letting Su Daqiang express his inner pain and sing “unravel”.

Want to try other characters besides Su Daqiang? Click the tutorial link to play with AI video face swap!

https://aistudio.baidu.com/aistudio/projectdetail/1048840

Face Generation

Want to try face swapping? But think cosmetic surgery is too expensive and risky? GAN can help you generate various faces, whether realistic or anime-style!

Image Transformation

Don’t blink, GAN can directly help you become “Shen Bi Ma Liang”, freely converting real images into “anime” images.

Such an interesting technical direction, are you also eager to dive into the research and application of GAN technology?

PaddleGAN Special Interest Group (SIG) Recruitment!

This time, the Paddle team will adopt a unique approach to bring together more enthusiasts interested in GAN technology to collaborate closely, building the PaddleGAN toolkit based on everyone’s actual interests and needs.

The concept of SIG is not unique to Paddle; it was first proposed and used by the ACM (Association for Computing Machinery) in 1961. Many top international open-source organizations use SIGs to allow members with specific interests to share, learn, or develop projects together. These members do not need to be in the same country/region or organization; as long as they share the same goals, they can learn, work, and play together.

PaddleGAN SIG is such an organization that gathers enthusiasts interested in GANs. Here, there are frontline developers from Baidu Paddle, senior engineers from Fortune 500 companies, and students from top universities at home and abroad.

Moreover, the renowned “composer most skilled at painting among programmers” – Dagu Spitzer, an independent artist, has also become an important member of PaddleGAN SIG. The restoration of the historical life scenes of old Beijing 100 years ago, which we discussed at the beginning, was achieved through Dagu’s application of GAN capabilities, and he has provided many valuable suggestions for enhancing PaddleGAN’s capabilities.

Dagu Spitzer

Hu Wengu, graduated from the School of Visual Arts (MFA in Computer Art). An independent artist, game developer, and freelance composer residing in New York. He is a popular science media author on AI technology, developer of “Eddy Purple” and “Convention Simulator”, and producer of the AI film “Old Beijing 100 Years Film Restoration”.

Dagu’s AI restoration series has attracted widespread attention, including “AI Restoration of Old Beijing 100 Years Ago”, “AI Restoration of a Fashion Show from 100 Years Ago”, “AI Restoration of the Terracotta Army from the Qin Dynasty”, and various AI technical restorations for the CCTV documentary series on the theme of resistance during the war, among others.

Among these, Dagu’s video “AI Restoration of Old Beijing 100 Years Ago” has been praised and shared by major media outlets such as CCTV News, People’s Daily, and Guangming Daily, sparking discussions among over 10 million netizens.

In addition to top independent artists, PaddleGAN SIG also includes experts from the industry:

Han Jie

Researcher at GE Research Center and Senior Engineer at GE China, currently the Chief Engineer at a startup. He has extensive experience in AI industrial applications, mainly researching AI robot detection and processing automation, and has led multiple projects implementing DL OCR, Detection, and Segmentation in industrial scenarios.

Ai Meng

Known as Ai Meng, a dreamer and fantasist programmer, a deeply experienced developer in Paddle, who is often called the “walking Paddle documentation”. He codes in C++ during the day and plays with AI after work. He has reproduced high-quality content such as StyleGAN2 and waifu2x based on PaddleGAN.

What are you waiting for? We welcome all experts and open-source enthusiasts to join us and explore the contributions that AI can bring!

You can:

Provide functional modules directly for PaddleGAN: such as models built natively on Paddle, prediction models converted to Paddle, datasets, papers, technical tutorials, etc.
Apply existing capabilities to develop creative applications: applications of various themes and scenarios across different industries are welcome, and we look forward to seeing everyone make GAN flourish in every field!
Publish academic papers (the Paddle team can participate directly).

As long as you are passionate about GAN and willing to share and contribute, SIG will also try to give back more to developers:

Join Us

Welcome to join the PaddleGAN WeChat group

Have the opportunity to become a member of a top global open-source project.
Participate in domestic and international enterprise, technology trends, and competition exchanges.
Have close contact, cooperation, and technical growth with professional developers and industry leaders.
Participate in promoting the Paddle community, including giving talks, sharing, and promoting projects.
Provide internship and job opportunities at Paddle.

Scan the QR code below to join the PaddleGAN SIG family after a technical assessment:

https://iwenjuan.baidu.com/?code=e4yfdo

We welcome your arrival here and look forward to your joining!

PaddleGAN is the Paddle Generative Adversarial Network toolkit, currently in rapid iterative development. We welcome everyone to provide valuable suggestions and support us.

GitHub link:

https://github.com/paddlepaddle/paddlegan

Gitee link:

https://gitee.com/paddlepaddle/PaddleGAN

PaddlePaddle is based on Baidu’s years of deep learning technology research and business applications, and is China’s first open-source, technologically advanced, and fully functional industrial-level deep learning platform, including the Paddle open-source platform and the Paddle enterprise version. The Paddle open-source platform includes core frameworks, basic model libraries, end-to-end development kits, and tool components, continuously open-sourcing core capabilities to provide a foundational base for industry, academia, and scientific innovation. The Paddle enterprise version enhances relevant features based on the Paddle open-source platform, including the zero-threshold AI development platform EasyDL and the fully functional AI development platform BML. EasyDL is mainly aimed at small and medium-sized enterprises, providing a zero-threshold, rich preset network and model, and a convenient and efficient development platform; BML is a fully functional, customizable, and deeply integrated development platform for large enterprises.

Related posts

Leave a Comment Cancel reply