Waymo LLC

Honda Motor Co., Ltd.

Alper Ayvaci

Yu-Han Chen

Bernd Heisele

Chen Wu

Ruichi Yu

Daniel Rudolf Maurer

Vasiliy Igorevich Karasev

Abhijit S Ogale

Hyman Murveit

Jonathon Shlens

Congcong Li

Tiffany L Chen

Zhinan Xu

Lo Po Tsui

Sean Rafferty

Noha Waheed Ahmed Radwan

Changchang Wu

Sheng Chen

Hang Yan

Kai-Chi Chan

Justin Thorsen

Wan-Yen Lo

Anelia Angelova

Chen Sun

Junhua Mao

Xu Chen

Kevin Patrick Murphy

Jiyang Gao

Kang Li

Hang Zhao

Henrik Kretzschmar

Nichola Abdo

Ruizhongtai Qi

Austin Charles Stone

Christopher John Sweeney

Bayram Safa Cicek

Dillon Cower

Justin Yu Zheng

Feiyu Chen

Kratarth Goel

Longlong Jing

Albert Zhao

Rico Jonschkowski

Jingjing Zheng

Yingying Zhu

Robert William Anderson

A method includes obtaining, by a processing device, input data derived from a set of sensors associated with an autonomous vehicle (AV). The input data includes camera data and radar data. The method further includes extracting, by the processing device from the input data, a plurality of sets of bird's-eye view (BEV) features. Each set of BEV features corresponds to a respective timestep. The method further includes generating, by the processing device from the plurality of sets of BEV features, an object flow for at least one object. Generating the object flow includes performing at least one of: multi-frame temporal aggregation or multi-frame dense motion estimation. The method further includes causing, by the processing device, a driving path of the AV to be modified in view of the object flow.

Multi-frame temporal aggregation and dense motion estimation for autonomous vehicles

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for predicting three-dimensional object locations from images. One of the methods includes obtaining a sequence of images that comprises, at each of a plurality of time steps, a respective image that was captured by a camera at the time step; generating, for each image in the sequence, respective pseudo-lidar features of a respective pseudo-lidar representation of a region in the image that has been determined to depict a first object; generating, for a particular image at a particular time step in the sequence, image patch features of the region in the particular image that has been determined to depict the first object; and generating, from the respective pseudo-lidar features and the image patch features, a prediction that characterizes a location of the first object in a three-dimensional coordinate system at the particular time step in the sequence.

Three-dimensional location prediction from images

The described aspects and implementations enable efficient detection and classification of objects with machine learning models that deploy a bird's-eye view representation and are trained using depth ground truth data. In one implementation, disclosed are system and techniques that include obtaining images, generating, using a first neural network (NN), feature vectors (FVs) and depth distributions pixels of images, wherein the first NN is trained using training images and a depth ground truth data for the training images. The techniques further include obtaining a feature tensor (FT) in view of the FVs and the depth distributions, and processing the obtained FTs, using a second NN, to identify one or more objects depicted in the images.

Object identification in bird's-eye view reference frame with explicit depth estimation co-training

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network to predict optical flow. One of the methods includes obtaining a batch of one or more training image pairs; for each of the pairs: processing the first training image and the second training image using the neural network to generate a final optical flow estimate; generating a cropped final optical flow estimate from the final optical flow estimate; and training the neural network using the cropped optical flow estimate.

Unsupervised training of optical flow estimation neural networks

Aspects of the disclosure provide for automatically generating labels for sensor data. For instance, first sensor data, for a vehicle may be identified. This first sensor data may have been captured by a first sensor of the vehicle at a first location during a first point in time and may be associated with a first label for an object. Second sensor data for the vehicle may be identified. The second sensor data may have been captured by a second sensor of the vehicle at a second location at a second point in time outside of the first point in time. The second location is different from the first location. A determination may be made as to whether the object is a static object. Based on the determination that the object is a static object, the first label may be used to automatically generate a second label for the second sensor data.

Automatic labeling of objects in sensor data

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a future occupancy prediction for a region of an environment. In one aspect, a method comprises: receiving sensor data generated by a sensor system of a vehicle that characterizes an environment in a vicinity of the vehicle as of a current time point, wherein the sensor data comprises a plurality of sensor samples characterizing the environment that were each captured at different time points; processing a network input comprising the sensor data using a neural network to generate an occupancy prediction output for a region of the environment, wherein: the occupancy prediction output characterizes, for one or more future intervals of time after the current time point, a respective likelihood that the region of the environment will be occupied by an agent in the environment during the future interval of time.

Occupancy prediction neural networks

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a neural network using contrastive learning. One of the methods includes obtaining a network input representing an environment; processing the network input using a first subnetwork of the neural network to generate a respective embedding for each location in the environment; processing the embeddings for each location in the environment using a second subnetwork of the neural network to generate a respective object prediction for each location; determining, for each of a plurality of pairs of the plurality of locations in the environment, whether the respective object predictions of the pair of locations characterize the same possible object or different possible objects; computing a respective contrastive loss value for each of the plurality of pairs of locations; and updating values for a plurality of parameters of the first subnetwork using the computed contrastive loss values.

Contrastive learning for object detection

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating motion detection based on optical flow. One of the methods includes obtaining a first image of a scene in an environment taken by an agent at a first time point and a second image of the scene at a second later time point. A point cloud characterizing the scene in the environment is obtained. A predicted optical flow is determined between the first image and the second image. A respective initial flow prediction for the point that represents motion of the point between the two time points is determined. A respective ego motion flow estimate for the point that represents a motion of the point induced by ego motion of the agent is determined. A respective motion prediction that indicates whether the point was static or in motion between the two time points is determined.

Optical flow based motion detection

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating object interaction predictions using a neural network. One of the methods includes obtaining a sensor input derived from data generated by one or more sensors that characterizes a scene. The sensor input is provided to an object interaction neural network. The object interaction neural network is configured to process the sensor input to generate a plurality of object interaction outputs. Each respective object interaction output includes main object information and interacting object information. The respective object interaction outputs corresponding to the plurality of regions in the sensor input are received as output of the object interaction neural network.

Interacted object detection neural network

Aspects of the disclosure provide for automatically generating labels for sensor data. For instance, first sensor data for a vehicle may be identified. This first sensor data may have been captured by a first sensor of the vehicle at a first location during a first point in time and may be associated with a first label for an object. Second sensor data for the vehicle may be identified. The second sensor data may have been captured by a second sensor of the vehicle at a second location at a second point in time outside of the first point in time. The second location is different from the first location. A determination may be made as to whether the object is a static object. Based on the determination that the object is a static object, the first label may be used to automatically generate a second label for the second sensor data.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining a location of a particular object relative to a vehicle. In one aspect, a method includes obtaining sensor data captured by one or more sensors of a vehicle. The sensor data is processed by a convolutional neural network to generate a sensor feature representation of the sensor data. Data is obtained which defines a particular spatial region in the sensor data that has been classified as including sensor data that characterizes the particular object. An object feature representation of the particular object is generated from a portion of the sensor feature representation corresponding to the particular spatial region. The object feature representation of the particular object is processed using a localization neural network to generate the location of the particular object relative to the vehicle.

Object localization using machine learning

Aspects of the disclosure relate to training and using a model for identifying actions of objects. For instance, LIDAR sensor data frames including an object bounding box corresponding to an object as well as an action label for the bounding box may be received. Each sensor frame is associated with a timestamp and is sequenced with respect to other sensor frames. Each given sensor data frame may be projected into a camera image of the object based on the timestamp associated with the given sensor data frame in order to provide fused data. The model may be trained using the fused data such that the model is configured to, in response to receiving fused data, the model outputs an action label for each object bounding box of the fused data. This output may then be used to control a vehicle in an autonomous driving mode.

Object action classification for autonomous vehicles

Disclosed are systems and methods for detecting moving objects. A computer-implemented method for detecting moving objects comprises obtaining a streaming video captured by a camera; extracting an input image sequence including a series of images from the streaming video; tracking point features and maintaining a set of point trajectories for at least one of the series of images; measuring a likelihood for each point trajectory to determine whether it belongs to a moving object using constraints from multi-view geometry; and determining a conditional random field (CRF) on an entire frame to obtain a moving object segmentation.

Method and system for moving object detection with single camera

Growing community of inventors

Santa Clara, CA, United States of America

Alper Ayvaci