Created AI that converts a 2D image into a 3D model in five seconds

Researchers from Adobe and the Australian National University have developed an algorithm that converts 2D images into 3D models in just 5 seconds. This breakthrough promises to revolutionize content creation in design and entertainment, but one must be wary of potential issues with blurry textures and copyright infringement.

Image source: Placidplace / Pixabay

The new algorithm, called Large Reconstruction Model (LRM), is a significant advance in quickly creating 3D models from a single image. The algorithm, trained on a large dataset of nearly a million 3D objects available in Objaverse and MVImgNet, demonstrated exceptional generalization ability across a wide range of images when creating 3D content.

Unlike previous models that were trained on limited datasets and focused on a single image category, LRM uses a 500 million parameter transformer architecture, which has become the basis of many developments in deep learning. This allows the AI ​​model to work effectively with a variety of image types, including real-world photos and visual collages created by other AI services such as DALL-E and Stable Diffusion.

Image source: yiconghong.me

Yicong Hong, lead author of the study, emphasized that LRM represents a significant breakthrough in the field of 3D reconstruction. The algorithm is capable of reproducing detailed geometry and complex textures such as wood grain while maintaining quality and accuracy.

The application of LRM can be truly diverse, from design and entertainment to the gaming industry. For designers and 3D artists, this means faster model creation, which is particularly important when developing video games and animations. In addition, the ability of ordinary users to use LRM promises to make 3D modeling accessible to more than just professionals. Now it will be possible to create detailed models even from photos taken with a smartphone.

However, LRM has its limitations, such as: B. Texture blurring in hidden areas of the image. It is also worth considering copyright issues, particularly around the use of images generated by other AI services. To demonstrate the capabilities of LRM, the researchers created Page with video demonstrations and interactive 3D models. This highlights the growing role of AI in the modern world and its potential impact on the future of the creative industries.