Google Inc.

Waymo LLC

Soeren Pirk

Anelia Angelova

Reza Mahjourian

Vincent Michael Casser

Seyed Mohammad Khansari Zadeh

Yunfei Bai

Alexander Toshev

Honglak Lee

Zhao Chen

Pierre Sermanet

Ariel Gordon

Yao Lu

Harrison Lynch

Xinchen Yan

Jasmine Hsu

Karol Hausman

Yuanzheng Gong

Suhani Vora

Jan Dlabal

Henrik Kretzschmar

Alex Zihao Zhu

Anthony Brohan

Methods, systems, and apparatus for processing inputs that include video frames using neural networks. In one aspect, a system comprises one or more computers configured to obtain a set of one or more training images and, for each training image, ground truth instance data that identifies, for each of one or more object instances, a corresponding region of the training image that depicts the object instance. For each training image in the set, the one or more computers process the training image using an instance segmentation neural network to generate an embedding output comprising a respective embedding for each of a plurality of output pixels. The one or more computers then train the instance segmentation neural network to minimize a loss function.

Training instance segmentation neural networks through contrastive learning

A system for generating a depth output for an image is described. The system receives input images that depict the same scene, each input image including one or more potential objects. The system generates, for each input image, a respective background image and processes the background images to generate a camera motion output that characterizes the motion of the camera between the input images. For each potential object, the system generates a respective object motion output for the potential object based on the input images and the camera motion output. The system processes a particular input image of the input images using a depth prediction neural network (NN) to generate a depth output for the particular input image, and updates the current values of parameters of the depth prediction NN based on the particular depth output, the camera motion output, and the object motion outputs for the potential objects.

Unsupervised depth prediction neural networks

Training and/or using a machine learning model for performing robotic tasks is disclosed herein. In many implementations, an environment-conditioned action sequence prediction model is used to determine a set of actions as well as a corresponding particular order for the actions for the robot to perform to complete the task. In many implementations, each action in the set of actions has a corresponding action network used to control the robot in performing the action.

Determining environment-conditioned action sequences for robotic tasks

Implementations relate to training a point cloud prediction model that can be utilized to process a single-view two-and-a-half-dimensional (2.5D) observation of an object, to generate a domain-invariant three-dimensional (3D) representation of the object. Implementations additionally or alternatively relate to utilizing the domain-invariant 3D representation to train a robotic manipulation policy model using, as at least part of the input to the robotic manipulation policy model during training, the domain-invariant 3D representations of simulated objects to be manipulated. Implementations additionally or alternatively relate to utilizing the trained robotic manipulation policy model in control of a robot based on output generated by processing generated domain-invariant 3D representations utilizing the robotic manipulation policy model.

Robotic manipulation using domain-invariant 3D representations predicted from 2.5D vision data

Training a machine learning model (e.g., a neural network model such as a convolutional neural network (CNN) model) so that, when trained, the model can be utilized in processing vision data (e.g., from a vision component of a robot), that captures an object, to generate a rich object-centric embedding for the vision data. The generated embedding can enable differentiation of even subtle variations of attributes of the object captured by the vision data.

Training a deep neural network model to generate rich object-centric embeddings of robotic vision data

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network using consistency measures. One of the methods includes processing a particular training example from a mediator training data set using a first neural network to generate a first output for a first machine learning task; processing the particular training example in the mediator training data set using each of one or more second neural networks, wherein each second neural network is configured to generate a second output for a respective second machine learning task; determining, for each second machine learning task, a consistency target output for the first machine learning task; determining, for each second machine learning task, an error between the first output and the consistency target output corresponding to the second machine learning task; and generating a parameter update for the first neural network from the determined errors.

Training neural networks using consistency measures

A method for generating a predicted segmentation map for potential objects in a future scene depicted in a future image is described. The method includes receiving input images that depict a same scene; processing a current input image to generate a segmentation map for potential objects in the current input image and a respective depth map; generating a point cloud for the current input image; processing the input images to generate, for each pair of two input images in the sequence, a respective ego-motion output that characterizes motion of the camera between the two input images; processing the ego-motion outputs to generate a future ego-motion output; processing the point cloud of the current input image and the future ego-motion output to generate a future point cloud; and processing the future point cloud to generate the predicted segmentation map for potential objects in the future scene depicted in the future image.

Future semantic segmentation prediction using 3D structure

Growing community of inventors

Palo Alto, CA, United States of America

Soeren Pirk