Google Inc.

Rohit Kumar Pandey

Sofien Bouaziz

Sean Ryan Francesco Fanello

Paul E Debevec

Ricardo Martin Brualla

Daniel Goldman

Shahram Izadi

Adarsh Prakash Murthy Kowdle

Jonathan James Taylor

Christoph Rhemann

Chloe LeGendre

Philip L Davidson

Graham Fyffe

Cem Keskin

Danhang Tang

Andrea Tagliasacchi

Mingsong Dou

Xueming Yu

Jessica Lynn Busch

Jason Angelo Dourgarian

Geoffrey Douglas Harvey

Moustafa Meshry

Pavlo Pidlypenskyi

Peter Christopher Lincoln

Rong Liu

Jonathan Tilton Barron

Ke Li

Ruofei Du

Yun-Ta Tsai

Sameh Khamis

Yinda Zhang

Shiqi Chen

Matthew Alun Brown

Julien Pascal Christophe Valentin

Noah Snavely

Christian Haene

Hugues Hoppe

Navin Sarma

Emily B Cooper

Marie Charisse White

Eric Aboussouan

Sergio Orts Escolano

Ivana Tosic Rodgers

Matthew Whalen

Deqing Sun

Wan-Chun Ma

Christina Tong

Thabo Beeler

Ryan Overbeck

Michael Milne

Anastasia Tkach

Kaiwen Guo

Shuoran Yang

Abhi Meka

Peter Barnum

Feitong Tan

Abhimitra Meka

Peter Joseph Denny

Xue Wang

Christian Häne

Matthew Brown

Example embodiments relate to techniques for volumetric performance capture with neural rendering. A technique may involve initially obtaining images that depict a subject from multiple viewpoints and under various lighting conditions using a light stage and depth data corresponding to the subject using infrared cameras. A neural network may extract features of the subject from the images based on the depth data and map the features into a texture space (e.g., the UV texture space). A neural renderer can be used to generate an output image depicting the subject from a target view such that illumination of the subject in the output image aligns with the target view. The neural render may resample the features of the subject from the texture space to an image space to generate the output image.

Volumetric performance capture with neural rendering

Apparatus and methods related to applying lighting models to images of objects are provided. An example method includes applying a geometry model to an input image to determine a surface orientation map indicative of a distribution of lighting on an object based on a surface geometry. The method further includes applying an environmental light estimation model to the input image to determine a direction of synthetic lighting to be applied to the input image. The method also includes applying, based on the surface orientation map and the direction of synthetic lighting, a light energy model to determine a quotient image indicative of an amount of light energy to be applied to each pixel of the input image. The method additionally includes enhancing, based on the quotient image, a portion of the input image. One or more neural networks can be trained to perform one or more of the aforementioned aspects.

Enhanced photo relighting based on machine learning models

Techniques of estimating lighting from portraits includes generating a lighting estimate from a single image of a face based on a machine learning (ML) system using multiple bidirectional reflection distribution functions (BRDFs) as a loss function. In some implementations, the ML system is trained using images of faces formed with HDR illumination computed from LDR imagery. The technical solution includes training a lighting estimation model in a supervised manner using a dataset of portraits and their corresponding ground truth illumination.

Learning illumination from diverse portraits

A lighting stage includes a plurality of lights that project alternating spherical color gradient illumination patterns onto an object or human performer at a predetermined frequency. The lighting stage also includes a plurality of cameras that capture images of an object or human performer corresponding to the alternating spherical color gradient illumination patterns. The lighting stage also includes a plurality of depth sensors that capture depth maps of the object or human performer at the predetermined frequency. The lighting stage also includes (or is associated with) one or more processors that implement a machine learning algorithm to produce a three-dimensional (3D) model of the object or human performer. The 3D model includes relighting parameters used to relight the 3D model under different lighting conditions.

Volumetric performance capture with relighting

Systems and methods are described for utilizing an image processing system with at least one processing device to perform operations including receiving a plurality of input images of a user, generating a three-dimensional mesh proxy based on a first set of features extracted from the plurality of input images and a second set of features extracted from the plurality of input images. The method may further include generating a neural texture based on a three-dimensional mesh proxy and the plurality of input images, generating a representation of the user including at least a neural texture, and sampling at least one portion of the neural texture from the three-dimensional mesh proxy. In response to providing the at least one sampled portion to a neural renderer, the method may include receiving, from the neural renderer, a synthesized image of the user that is previously not captured by the image processing system.

Few-shot synthesis of talking heads

Systems and methods for training models to predict dense correspondences across images such as human images. A model may be trained using synthetic training data created from one or more 3D computer models of a subject. In addition, one or more geodesic distances derived from the surfaces of one or more of the 3D models may be used to generate one or more loss values, which may in turn be used in modifying the model's parameters during training.

Systems and methods for training models to predict dense correspondences in images using geodesic distances

Techniques of tracking a user's gaze includes identifying a region of a display at which a gaze of a user is directed, the region including a plurality of pixels. By determining a region rather than a point, when the regions correspond to elements of a user interface, the improved technique enables a system to activate the element to which a determined region is selected. In some implementations, the system makes the determination using a classification engine including a convolutional neural network; such an engine takes as input images of the user's eye and outputs a list of probabilities that the gaze is directed to each of the regions.

Eye gaze classification

Systems and methods are described for generating a plurality of three-dimensional (3D) proxy geometries of an object, generating, based on the plurality of 3D proxy geometries, a plurality of neural textures of the object, the neural textures defining a plurality of different shapes and appearances representing the object, providing the plurality of neural textures to a neural renderer, receiving, from the neural renderer and based on the plurality of neural textures, a color image and an alpha mask representing an opacity of at least a portion of the object, and generating a composite image based on the pose, the color image, and the alpha mask.

Generative latent textured proxies for object category modeling

A method includes receiving a first image including color data and depth data, determining a viewpoint associated with an augmented reality (AR) and/or virtual reality (VR) display displaying a second image, receiving at least one calibration image including an object in the first image, the object being in a different pose as compared to a pose of the object in the first image, and generating the second image based on the first image, the viewpoint and the at least one calibration image.

Volumetric capture of objects with a single RGBD camera

According to an aspect, a method for neural rerendering includes obtaining a three-dimensional (3D) model representing a scene of a physical space, where the 3D model is constructed from a collection of input images, rendering an image data buffer from the 3D model according to a viewpoint, where the image data buffer represents a reconstructed image from the 3D model, receiving, by a neural rerendering network, the image data buffer, receiving, by the neural rerendering network, an appearance code representing an appearance condition, and transforming, by the neural rerendering network, the image data buffer into a rerendered image with the viewpoint of the image data buffer and the appearance condition specified by the appearance code.

Neural rerendering from 3D models

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

In at least one aspect, a method can include generating a respective set of training set of images for each label in a handedness model by: receiving the label at an image capturing device, obtaining a set of captured images by recording a pass-through image of a user placing a target object within an overlay of a bounding area animation, the target object corresponding with the label, and associating the label with each image in the set of captured images. The method includes training, using the training images, the handedness model to provide a correct label for an input image.

Methods of determining handedness for virtual controllers

Growing community of inventors

Mountain View, CA, United States of America

Rohit Kumar Pandey