Apple Machine study group, along with scientists from universities in Nanjing and Hong Kong, have developed a new model of artificial intelligence (AI) called Matrix3D. She can recreate objects and scenes with only three dimensions of three two -way images.
Matrix3D belongs to large photogrammetric models. Unlike traditional methods, in which each stage – from determining the depth assessment angle – requires individual models, Matrix3D combines all processes in a system. This makes building 3D models faster and more accurate.
During the training, the model used a strategy to hide the data in which the random parts of the input information were hidden. This approach makes the system end, the paragraphs are missing, allowing it to operate even with incomplete or small data sets. A similar principle has been applied in early transformer models that are founded for the development of II systems such as chatgpt.
Matrix3D can be especially useful for additional actual and virtual practical technologies, including Apple Vision Pro headphones. Thanks to the ability to create the official 3D scenes of all three images, such a system can be used in different fields-from design and architecture for games and digital duplication.
© Apple