This AI Research from Adobe Proposes a Large Reconstruction Model (LRM) that Predicts the 3D Model of an Object from a Single Input Image within 5 Seconds

Many researchers have envisioned a world where any 2D image can be instantaneously converted into a 3D model. Research in this area has been mostly motivated by the desire to find a generic and efficient method of achieving this long-standing objective, with potential applications spanning industrial design, animation, gaming, and augmented reality/virtual reality.  Early learning-based approaches typically perform well on certain categories, using the category data before inferring the overall shape because of the inherent ambiguity of 3D geometry in a single glance. Recent studies have been motivated by recent developments in image generation, such as DALL-E and Stable Diffusion,

This is a companion discussion topic for the original entry at