Meta Platforms Technologies, LLC

Michael Goesele

Richard Andrew Newcombe

Steven John Lovegrove

Michael Zollhoefer

Daniel Andersen

Simon Gareth Green

Changil Kim

Tianye Li

Tanner Schmidt

Grace Elizabeth Kuo

Zhaoyang Lv

Miroslava Slavcheva

Christoph Lassner

Yujia Chen

Zheng Shi

In particular embodiments, a computing system may capture a first image of a scene using a first camera of an artificial reality device. The system may capture a second image of the scene using a second camera and one or more optical elements of the artificial reality device. The second image may include an overlapping portion of multiple shifted copies of the scene. The system may generate an upsampled first image by applying a particular sampling technique to the first image. The system may generate a tiled image comprising a plurality of repeated second images by applying a tiling process to the second image. The system may generate an initial output image by processing the upsampled first image and the tiled image using a machine learning model. The system may generate a final output image by normalizing the initial output image using the upsampled first image.

Generating super resolution images using duo-camera artificial reality device

In one embodiment, a method includes initializing latent codes respectively associated with times associated with frames in a training video of a scene captured by a camera. For each of the frames, a system (1) generates rendered pixel values for a set of pixels in the frame by querying NeRF using the latent code associated with the frame, a camera viewpoint associated with the frame, and ray directions associated with the set of pixels, and (2) updates the latent code associated with the frame and the NeRF based on comparisons between the rendered pixel values and original pixel values for the set of pixels. Once trained, the system renders output frames for an output video of the scene, wherein each output frame is rendered by querying the updated NeRF using one of the updated latent codes corresponding to a desired time associated with the output frame.

Neural 3D video synthesis

Growing community of inventors

Woodinville, WA, United States of America

Michael Goesele