Turning photos into spatial models has long been considered the holy grail of computer vision and computer graphics, according to David Luebke, Nvidia’s vice president of graphics research. Nvidia seems to be getting closer to 3D MoMa rendering software now. The tech giant promises that the AI-based tool should help architects, designers, artists, and other professionals easily load objects into their software.
The shape used for this is a triangle mesh used in 3D design and games. Therefore, the format is compatible with commonly used editing software and can be used for all kinds of editing after uploading, such as scaling, distorting, and adjusting texture and color.
The 3D model is created on the basis of photographs of the object in question. The material consists of a 2D image superimposed on the model and the lighting is based on photographs. With a single Tensor core on an Nvidia GPU, a model can be built in an hour.
The manufacturer demonstrates this with five different musical instruments, of which 100 have been photographed. In the shared video, it can be seen that they react to the light when it comes from a different angle. The article “Extracting Triangle 3D Models, Materials, and Lighting from Images” goes into further detail on how the software works.
Source: HotHardware
Source: Hardware Info
