Microsoft has released a diffusion model that can build a 3D avatar from a single photo of a person

News Report Technology

In Brief

A single 2D image of a person’s face can be used to generate a 3D avatar using the 3D Avatar Diffusion machine learning technique.

It can be used to provide a realistic 3D view of the person for gaming or other uses, or to provide a virtual reality (VR) or augmented reality (AR) experience.


The Trust Project is a worldwide group of news organizations working to establish transparency standards.

The 3D Avatar Diffusion is a machine learning algorithm that can take a single 2D image of a human face and create a three-dimensional (3D) avatar. The avatar can then be used to create a virtual reality (VR) or augmented reality (AR) experience or to simply provide a realistic 3D view of the person for gaming or other purposes.

The diffusion model was developed by a team of researchers at Microsoft Research and is described in a paper published in the journal arXiv.

Microsoft has released a diffusion model that can build a 3D avatar from a single photo of a person

The 3D Avatar Diffusion is based on a type of machine-learning algorithm called a diffusion model. Diffusion models are generative models, which means they can generate new data that is similar to the training data. Diffusion models have been used before to generate 3D images from 2D images, but the ADM is the first diffusion model that can generate a realistic 3D avatar from a single 2D image.

To train the model, the researchers used a dataset of over 200,000 3D face models. The dataset included a wide variety of faces with different skin tones, hairstyles, and facial features. The ADM was then able to learn the relationship between the 2D image and the 3D face model and generate a realistic 3D avatar from a single 2D image.

The model can also be used to generate an avatar from a photo that has been taken from a different angle

For the personalized 3D avatar, Rodin model offers text-guided manipulation. Natural language editing is an intuitive way to change many different 3D avatar features.

This study proposes a 3D generative model that automatically creates 3D digital avatars that are represented as neural radiance fields using diffusion models. Because of the prohibitive memory and processing requirements associated with 3D, creating the rich features necessary for high-quality avatars is a huge issue. Developers suggest the roll-out diffusion network (Rodin) address this issue.

In terms of gender, age, race, expression, facial accessories, etc., the model exhibits outstanding generational diversity. ​

This network rolls out numerous 2D feature maps of a neural radiance field into a single 2D feature plane, where the model then executes 3D-aware diffusion. The Rodin model uses 3D-aware convolution, which attends to projected features in the 2D feature plane according to their original relationship in 3D, to provide the much-needed computational efficiency while maintaining the integrity of diffusion in 3D.

Read more about AI:

Disclaimer

Any data, text, or other content on this page is provided as general market information and not as investment advice. Past performance is not necessarily an indicator of future results.

Damir Yalalov

Damir is the Editor/SEO/Product Lead at mpost.io. He is most interested in SecureTech, Blockchain, and FinTech startups. Damir earned a bachelor's degree in physics.

Follow Author

More Articles
🗞 Metaverse Newsletter
👾 Follow us
  YouTube Icon     YouTube Icon     YouTube Icon     YouTube Icon