Google Inc.

Julien Pascal Christophe Valentin

Shahram Izadi

Adarsh Prakash Murthy Kowdle

Christoph Rhemann

Vladimir Tankovich

Philip L Davidson

Sean Ryan Fanello

Jonathan James Taylor

Paul E Debevec

Sofien Bouaziz

Sean Ryan Francesco Fanello

Rohit Kumar Pandey

Graham Fyffe

Cem Keskin

David Kim

Sameh Khamis

Danhang Tang

Andrea Tagliasacchi

Mingsong Dou

Xueming Yu

Jessica Lynn Busch

Jason Angelo Dourgarian

Matthew Whalen

Peter Christopher Lincoln

Geoffrey Douglas Harvey

Peter Joseph Denny

Christian Häne

Abhimitra Meka

Values of pixels in an image are mapped to a binary space using a first function that preserves characteristics of values of the pixels. Labels are iteratively assigned to the pixels in the image in parallel based on a second function. The label assigned to each pixel is determined based on values of a set of nearest-neighbor pixels. The first function is trained to map values of pixels in a set of training images to the binary space and the second function is trained to assign labels to the pixels in the set of training images. Considering only the nearest neighbors in the inference scheme results in a computational complexity that is independent of the size of the solution space and produces sufficient approximations of the true distribution when the solution for each pixel is most likely found in a small subset of the set of potential solutions.

Fully parallel, low complexity approach to solving computer vision problems

Methods, systems, and media for relighting images using predicted deep reflectance fields are provided. In some embodiments, the method comprises: identifying a group of training samples, wherein each training sample includes (i) a group of one-light-at-a-time (OLAT) images that have each been captured when one light of a plurality of lights arranged on a lighting structure has been activated, (ii) a group of spherical color gradient images that have each been captured when the plurality of lights arranged on the lighting structure have been activated to each emit a particular color, and (iii) a lighting direction, wherein each image in the group of OLAT images and each of the spherical color gradient images are an image of a subject, and wherein the lighting direction indicates a relative orientation of a light to the subject; training a convolutional neural network using the group of training samples, wherein training the convolutional neural network comprises: for each training iteration in a series of training iterations and for each training sample in the group of training samples: generating an output predicted image, wherein the output predicted image is a representation of the subject associated with the training sample with lighting from the lighting direction associated with the training sample; identifying a ground-truth OLAT image included in the group of OLAT images for the training sample that corresponds to the lighting direction for the training sample; calculating a loss that indicates a perceptual difference between the output predicted image and the identified ground-truth OLAT image; and updating parameters of the convolutional neural network based on the calculated loss; identifying a test sample that includes a second group of spherical color gradient images and a second lighting direction; and generating a relit image of the subject included in each of the second group of spherical color gradient images with lighting from the second lighting direction using the trained convolutional neural network.

Methods, systems, and media for relighting images using predicted deep reflectance fields

An electronic device estimates a depth map of an environment based on stereo depth images captured by depth cameras having exposure times that are offset from each other in conjunction with illuminators pulsing illumination patterns into the environment. A processor of the electronic device matches small sections of the depth images from the cameras to each other and to corresponding patches of immediately preceding depth images (e.g., a spatio-temporal image patch 'cube'). The processor computes a matching cost for each spatio-temporal image patch cube by converting each spatio-temporal image patch into binary codes and defining a cost function between two stereo image patches as the difference between the binary codes. The processor minimizes the matching cost to generate a disparity map, and optimizes the disparity map by rejecting outliers using a decision tree with learned pixel offsets and refining subpixels to generate a depth map of the environment.

System and method for active stereo depth sensing

An electronic device estimates a pose of a face by fitting a generative face model mesh to a depth map based on vertices of the face model mesh that are estimated to be visible from the point of view of a depth camera. A face tracking module of the electronic device receives a depth image of a face from a depth camera and generates a depth map of the face based on the depth image. The face tracking module identifies a pose of the face by fitting a face model mesh to the pixels of a depth map that correspond to the vertices of the face model mesh that are estimated to be visible from the point of view of the depth camera.

High speed, high-fidelity face tracking

A first and second image of a scene are captured. Each of a plurality of pixels in the first image is associated with a disparity value. An image patch associated with each of the plurality of pixels of the first image and the second image is mapped into a binary vector. Thus, values of pixels in an image are mapped to a binary space using a function that preserves characteristics of values of the pixels. The difference between the binary vector associated with each of the plurality of pixels of the first image and its corresponding binary vector in the second image designated by the disparity value associated with each of the plurality of pixels of the first image is determined. Based on the determined difference between binary vectors, correspondence between the plurality of pixels of the first image and the second image is established.

Learning-based matching for active stereo systems

Growing community of inventors

Mountain View, CA, United States of America

Julien Pascal Christophe Valentin