3D Scene From a Single Image

⊲ Mixed News ⊳ Meta’s generative AI research team unveils a new method for generating 3D assets from a single 2D image, significantly reducing the time and expertise usually required. By combining two AI models – one for predicting depth and another for rendering texture and structure – they can rapidly convert photos and rendered using Gaussian splatting into realistic 3D objects. The technique could support applications in gaming, AR/VR, and the wider metaverse. However, it’s still under development and it is difficult to extend the navigable area beyond two meters.
⊲ Image – Meta ⊳