1. PixelPlaygroundHome
  2. Web3 News

Alibaba Composer: Unleashing the Creativity of AIGC

• Web3 News • 149 views

On April 12th, at the 2023 Web3 Carnival in Hong Kong, Zhao Deli, Director of the Basic Visual Team at Alibaba Dharma Institute, delivered a keynote speech titled \”Alibaba Composer

Alibaba Composer: Unleashing the Creativity of AIGC

On April 12th, at the 2023 Web3 Carnival in Hong Kong, Zhao Deli, Director of the Basic Visual Team at Alibaba Dharma Institute, delivered a keynote speech titled “Alibaba Composer, Unleashing the Creativity of AIGC” during the “New Trends in Web3 Technology: Artificial Intelligence Generated Content (AIGC) and Privacy Computing Technology” event. She stated that the future trends of the AI industry, including the development of AI algorithms in vertical industries, will continue to be deeply cultivated, The emergence of open source platforms (such as the ModelScope model launched by the Dharma Academy), and the unification of underlying algorithms. Our roadmap has gone through three stages in Text to Image: Foundation Models (Composer 1.0), Customized Generation, and Controllable Generation (Composer 2.0). This year, we have released ControlNet, Composer 2.0, and T2I Adapter. We have built a “Tongyi Wanxiang” product based on Alibaba Cloud, which will be open to everyone at the end of this month.

Director of the Basic Visual Team at Alibaba Dharma Academy: The text generated image product “Tongyi Wanxiang” will gradually be launched by the end of this month

Introduction

On April 12th, at the 2023 Web3 Carnival in Hong Kong, Zhao Deli, Director of the Basic Visual Team at Alibaba Dharma Institute, delivered a keynote speech titled “Alibaba Composer, Unleashing the Creativity of AIGC” during the “New Trends in Web3 Technology: Artificial Intelligence Generated Content (AIGC) and Privacy Computing Technology” event. In her speech, she shared her insights on the future trends of the AI industry, including the development of AI algorithms in vertical industries, the emergence of open source platforms, and the unification of underlying algorithms. She also discussed Alibaba’s roadmap for Text to Image and the recent releases of ControlNet, Composer 2.0, and T2I Adapter.

Alibaba’s Roadmap for Text to Image

Zhao Deli explained that Alibaba’s roadmap for Text to Image has gone through three stages: Foundation Models (Composer 1.0), Customized Generation, and Controllable Generation (Composer 2.0).

Foundation Models (Composer 1.0)

At the foundation stage, Alibaba developed Composer 1.0, which is a deep learning algorithm that can generate high-quality images based on textual descriptions. The algorithm was trained on millions of images and their corresponding textual descriptions, enabling it to understand and generate images based on a wide range of inputs.

Customized Generation

At the next stage, Alibaba focused on customized generation, where the algorithm can generate images that match the specific requirements of users. This involves training the algorithm on specific datasets relevant to a particular industry, enabling it to generate images that are tailored to the specific needs of that industry. This stage is important for the development of AI algorithms in vertical industries.

Controllable Generation (Composer 2.0)

At the final stage, Alibaba developed Composer 2.0, which allows users to manipulate generated images in real-time using a set of controls. This enables users to tweak the features of the image to match their exact requirements, providing a higher degree of control and flexibility in the generation of images.

Recent Releases

Zhao Deli also discussed the recent releases of ControlNet, Composer 2.0, and T2I Adapter. These releases demonstrate Alibaba’s commitment to continued innovation and development in the AI industry.

ControlNet

ControlNet is an open-source platform that enables developers to use a variety of deep learning algorithms to build novel neural networks. It provides a set of pre-trained and customizable models that can be used for a wide range of applications, including image classification, object detection, and semantic segmentation.

T2I Adapter

T2I Adapter is an algorithm that enables text-to-image generation in real-time. It is designed to work seamlessly with Alibaba’s deep learning algorithms, allowing users to generate high-quality images based on textual descriptions.

“Tongyi Wanxiang” Product

Zhao Deli also announced the launch of a new product called “Tongyi Wanxiang” based on Alibaba Cloud. The product will be open to everyone at the end of this month and will enable users to generate high-quality images based on textual descriptions using Composer 2.0 and T2I Adapter.

Conclusion

Alibaba’s continued innovation and development in the AI industry demonstrate its commitment to providing valuable solutions to real-world problems. With the development of Text to Image and the recent releases of ControlNet, Composer 2.0, and T2I Adapter, Alibaba is well-positioned to continue leading the industry in AI-generated content.

FAQs

1. What is the purpose of Alibaba Composer?
Alibaba Composer is a deep learning algorithm that can generate high-quality images based on textual descriptions. It is designed to provide a higher degree of control and flexibility in the generation of images.
2. What are the recent releases of Alibaba in the AI industry?
Alibaba has recently released ControlNet, Composer 2.0, and T2I Adapter. These releases demonstrate its commitment to continued innovation and development in the AI industry.
3. What is “Tongyi Wanxiang” product?
“Tongyi Wanxiang” is a new product based on Alibaba Cloud that enables users to generate high-quality images based on textual descriptions using Composer 2.0 and T2I Adapter. It will be open to everyone at the end of this month.

This article and pictures are from the Internet and do not represent Fpips's position. If you infringe, please contact us to delete:https://www.fpips.com/15124/

It is strongly recommended that you study, review, analyze and verify the content independently, use the relevant data and content carefully, and bear all risks arising therefrom.

Like (0)
Previous 04/12/2023 08:19
Next 04/12/2023 08:26

Recommended articles