Microsoft Technology Licensing, LLC

Michelle Brook

Szymon Piotr Stachniak

Hendrik Mark Langerak

Yuri Pekelny

Michael Bleyer

William Douglas Guyman

Marc Pollefeys

Ali Osman Ulusoy

Muhammad Jabir Kapasi

Silvano Galliani

Rahul Sawhney

The described implementations relate to images and depth information and generating useful information from the images and depth information. One example can identify planes in a semantically-labeled 3D voxel representation of a scene. The example can infer missing information by extending planes associated with structural elements of the scene. The example can also generate a watertight manifold representation of the scene at least in part from the inferred missing information.

Deriving information from images

Improved techniques for performing object segmentation are disclosed. Surface reconstruction (SR) data corresponding to an environment is accessed. This SR data is used to generate a detailed three-dimensional (3D) representation of the environment. The SR data is also used to infer a high-level 3D structural representation of the environment. The high-level 3D structural representation is inferred using machine learning that is performed on the surface reconstruction data to identify a structure of the environment. The high-level 3D structural representation is then cut from the detailed 3D representation. This cutting process generates a clutter mesh comprising objects that remain after the cut and that are distinct from the structure. Object segmentation is then performed on the remaining objects to identify those objects.

Object segmentation using high-level structural meshes

A computer system is provided that includes a camera device and a processor configured to receive scene data captured by the camera device for a three-dimensional environment that includes one or more physical objects, generate a geometric representation of the scene data, process the scene data using an artificial intelligence machine learning model that outputs object boundary data and object labels, augment the geometric representation with the object boundary data and the object labels, and identify the one or more physical objects based on the augmented geometric representation of the three-dimensional environment. For each identified physical object, the processor is configured to generate an associated virtual object that is fit to one or more geometric characteristics of that identified physical object. The processor is further configured to track each identified physical object and associated virtual object across successive updates to the scene data.

Object permanence in surface reconstruction

An augmented reality device includes a logic machine and a storage machine holding instructions executable by the logic machine to, for one or more real-world surfaces represented in a three-dimensional representation of a real-world environment of the augmented reality device, fit a virtual two-dimensional plane to the real-world surface. A request to place a virtual three-dimensional object on the real-world surface is received. For each of a plurality of candidate placement locations on the virtual two-dimensional plane, the candidate placement location is evaluated as a valid placement location or an invalid placement location for the virtual three-dimensional object. An invalidation mask is generated that defines the valid and invalid placement locations on the virtual two-dimensional plane.

Virtual object placement for augmented reality

A computer-implemented technique is described herein for discriminatively apprising the user of the existence of some, but not necessarily all, physical objects in a physical environment in the course of the user's interaction with a virtual environment. In operation, the technique allows the user to selectively identify one or more objects-of-interest, such as people, walls, computing devices, etc. The technique then uses a scene analysis component to automatically detect the presence of the selected objects-of-interest in the physical environment, while the user interacts with the virtual environment. The technique provides alert information which notifies the user of the existence of any objects-of-interest that it detects. By virtue of the above-summarized strategy, the technique can apprise the user of objects-of-interest in the user's vicinity without cluttering the user's virtual experience with extraneous information pertaining to the physical environment.

Growing community of inventors

Redmond, WA, United States of America

Michelle Brook