Nvidia Corporation

Minwoo Park

David Nister

Sangmin Oh

Neda Cvijetic

Yue Wu

Junghyun Kwon

Berta Rodriguez Hervas

Hae-Jong Seo

Pekka Janis

Yilin Yang

Zhaoting Ye

Trung Pham

Cheng-Chieh Yang

Bala Siva Sashank Jujjavarapu

Hang Dou

Xiaolin Lin

Gang Pan

Tommi Koivisto

Mehmet K Kocamaz

Davide Onofrio

Kang Wang

Daniel Herrera Castro

Tae Eun Choe

John R Zedlewski

Zuoguan Wang

Gordon Grigor

Miguel Sainz

Chia-Chih Chen

Abhishek Bajpayee

Igor Tryndin

Yifang Xu

Carolina Parada

Vijay Chintalapudi

Xin Liu

Mehdi Sajjadi Mohammadabadi

Ozan Tonkal

Sayed Mehdi Sajjadi Mohammadabadi

Jan Nikolaus Fritsch

I-Kuei Chen

Michael Grabner

David Weikersdorfer

Emilie Lucie Eloïse Wirbel

Arthur Henry Rajala

Aman Jhunjhunwala

Qian Lin

Gyeong Woo Cheon

Kexuan Zou

Bor-Jeng Chen

Xudong Chen

Eric Jonathan Yuan

Xin Tong

Yining Deng

Jagadeesh Sankaran

Sean Midthun Pieper

Wanli Jiang

Eric Dujardin

Dong Zhang

Eric Viscito

Baris Evrim Demiroz

Nizar Assaf

Siyi Li

Yichun Shen

Jincheng Li

Neeraj Sajjan

Daniel Per Olof Svensson

Hsin-I Chen

Pengfei Hao

Soohwan Kim

Chethan Ningaraju

Hao Tsui

Yongqing Liang

Bala Siva Jujjavarapu

In various examples, techniques for multi-dimensional tracking of objects using two-dimensional (2D) sensor data are described. Systems and methods may use first image data to determine a first 2D detected location and a first three-dimensional (3D) detected location of an object. The systems and methods may then determine a 2D estimated location using the first 2D detected location and a 3D estimated location using the first 3D detected location. The systems and methods may use second image data to determine a second 2D detected location and a second 3D detected location of a detected object, and may then determine that the object corresponds to the detected object using the 2D estimated location, the 3D estimated location, the second 2D detected location, and the second 3D detected location. The systems and method then generate, modify, delete, or otherwise update an object track that includes 2D state information and 3D state information.

Joint 2D and 3D object tracking for autonomous systems and applications

In various examples, live perception from sensors of a vehicle may be leveraged to detect and classify intersections in an environment of a vehicle in real-time or near real-time. For example, a deep neural network (DNN) may be trained to compute various outputs—such as bounding box coordinates for intersections, intersection coverage maps corresponding to the bounding boxes, intersection attributes, distances to intersections, and/or distance coverage maps associated with the intersections. The outputs may be decoded and/or post-processed to determine final locations of, distances to, and/or attributes of the detected intersections.

Intersection detection and classification in autonomous machine applications

In various examples, one or more deep neural networks (DNNs) are executed to regress on control points of a curve, and the control points may be used to perform a curve fitting operation—e.g., Bezier curve fitting—to identify landmark locations and geometries in an environment. The outputs of the DNN(s) may thus indicate the two-dimensional (2D) image-space and/or three-dimensional (3D) world-space control point locations, and post-processing techniques—such as clustering and temporal smoothing—may be executed to determine landmark locations and poses with precision and in real-time. As a result, reconstructed curves corresponding to the landmarks—e.g., lane line, road boundary line, crosswalk, pole, text, etc.—may be used by a vehicle to perform one or more operations for navigating an environment.

Landmark detection using curve fitting for autonomous driving applications

In various examples, systems and methods are disclosed that detect hazards on a roadway by identifying discontinuities between pixels on a depth map. For example, two synchronized stereo cameras mounted on an ego-machine may generate images that may be used extract depth or disparity information. Because a hazard's height may cause an occlusion of the driving surface behind the hazard from a perspective of a camera(s), a discontinuity in disparity values may indicate the presence of a hazard. For example, the system may analyze pairs of pixels on the depth map and, when the system determines that a disparity between a pair of pixels satisfies a disparity threshold, the system may identify the pixel nearest the ego-machine as a hazard pixel.

Multi-view geometry-based hazard detection for autonomous systems and applications

In various examples, one or more output channels of a deep neural network (DNN) may be used to determine assignments of obstacles to paths. To increase the accuracy of the DNN, the input to the DNN may include an input image, one or more representations of path locations, and/or one or more representations of obstacle locations. The system may thus repurpose previously computed information—e.g., obstacle locations, path locations, etc.—from other operations of the system, and use them to generate more detailed inputs for the DNN to increase accuracy of the obstacle to path assignments. Once the output channels are computed using the DNN, computed bounding shapes for the objects may be compared to the outputs to determine the path assignments for each object. Additionally, a machine may perform control operations based at least on the path assignments.

Obstacle to path assignment and performance of control operations based on assignment for autonomous systems and applications

In various examples, techniques for sensor-fusion based object detection and/or free-space detection using ultrasonic sensors are described. Systems may receive sensor data generated using one or more types of sensors of a machine. In some examples, the systems may then process at least a portion of the sensor data to generate input data, where the input data represents one or more locations of one or more objects within an environment. The systems may then input at least a portion of the sensor data and/or at least a portion of the input data into one or more neural networks that are trained to output one or more maps or other output representations associated with the environment. In some examples, the map(s) may include a height, an occupancy, and/or height/occupancy map generated, e.g., from a birds-eye-view perspective. The machine may use these outputs to perform one or more operations.

Sensor fusion using ultrasonic sensors for autonomous systems and applications

In various examples, a multi-sensor fusion machine learning model—such as a deep neural network (DNN)—may be deployed to fuse data from a plurality of individual machine learning models. As such, the multi-sensor fusion network may use outputs from a plurality of machine learning models as input to generate a fused output that represents data from fields of view or sensory fields of each of the sensors supplying the machine learning models, while accounting for learned associations between boundary or overlap regions of the various fields of view of the source sensors. In this way, the fused output may be less likely to include duplicate, inaccurate, or noisy data with respect to objects or features in the environment, as the fusion network may be trained to account for multiple instances of a same object appearing in different input representations.

Sensor fusion for autonomous machine applications using machine learning

In various examples, live perception from sensors of a vehicle may be leveraged to detect and classify intersection contention areas in an environment of a vehicle in real-time or near real-time. For example, a deep neural network (DNN) may be trained to compute outputs—such as signed distance functions—that may correspond to locations of boundaries delineating intersection contention areas. The signed distance functions may be decoded and/or post-processed to determine instance segmentation masks representing locations and classifications of intersection areas or regions. The locations of the intersections areas or regions may be generated in image-space and converted to world-space coordinates to aid an autonomous or semi-autonomous vehicle in navigating intersections according to rules of the road, traffic priority considerations, and/or the like.

Intersection region detection and classification for autonomous machine applications

In various examples, a three-dimensional (3D) intersection structure may be predicted using a deep neural network (DNN) based on processing two-dimensional (2D) input data. To train the DNN to accurately predict 3D intersection structures from 2D inputs, the DNN may be trained using a first loss function that compares 3D outputs of the DNN—after conversion to 2D space—to 2D ground truth data and a second loss function that analyzes the 3D predictions of the DNN in view of one or more geometric constraints—e.g., geometric knowledge of intersections may be used to penalize predictions of the DNN that do not align with known intersection and/or road structure geometries. As such, live perception of an autonomous or semi-autonomous vehicle may be used by the DNN to detect 3D locations of intersection structures from 2D inputs.

Three-dimensional intersection structure prediction for autonomous driving applications

In various examples, sensor data representative of an image of a field of view of a vehicle sensor may be received and the sensor data may be applied to a machine learning model. The machine learning model may compute a segmentation mask representative of portions of the image corresponding to lane markings of the driving surface of the vehicle. Analysis of the segmentation mask may be performed to determine lane marking types, and lane boundaries may be generated by performing curve fitting on the lane markings corresponding to each of the lane marking types. The data representative of the lane boundaries may then be sent to a component of the vehicle for use in navigating the vehicle through the driving surface.

Real-time detection of lanes and boundaries by autonomous vehicles

In various examples, systems and methods are disclosed that preserve rich spatial information from an input resolution of a machine learning model to regress on lines in an input image. The machine learning model may be trained to predict, in deployment, distances for each pixel of the input image at an input resolution to a line pixel determined to correspond to a line in the input image. The machine learning model may further be trained to predict angles and label classes of the line. An embedding algorithm may be used to train the machine learning model to predict clusters of line pixels that each correspond to a respective line in the input image. In deployment, the predictions of the machine learning model may be used as an aid for understanding the surrounding environment—e.g., for updating a world model—in a variety of autonomous machine applications.

Regression-based line detection for autonomous driving machines

In various examples, apparatuses, systems, and techniques to perform offline image signal processing of source image data to generate target image data. In at least one embodiment, data collection using exposure and calibration setting of an image sensor is performed to generate source image data, which is then processed by using offline image signal processing to generate target data.

High dynamic range image processing with fixed calibration settings

In various examples, perception-based parking assistance systems and methods for an ego-machine are presented. Example embodiments may determine a location of a real-world parking strip relative to an ego-machine and an associated parking rule for the parking strip. A virtual parking strip and one or more virtual parking signs may be generated based at least in part on one or more detected features in an environment of the ego-machine and a tracked motion of the ego machine, and the virtual parking strip may be used to track parking strip locations and associated parking rules. The virtual parking strips and associated rules may be relied upon by an ego-machine to determine parking locations and/or to navigate into a suitable parking spot.

Perception-based parking assistance for autonomous machine systems and applications

In various examples, to support training a deep neural network (DNN) to predict a dense representation of a 3D surface structure of interest, a training dataset is generated using a simulated environment. For example, a simulation may be run to simulate a virtual world or environment, render frames of virtual sensor data (e.g., images), and generate corresponding depth maps and segmentation masks (identifying a component of the simulated environment such as a road). To generate input training data, 3D structure estimation may be performed on a rendered frame to generate a representation of a 3D surface structure of the road. To generate corresponding ground truth training data, a corresponding depth map and segmentation mask may be used to generate a dense representation of the 3D surface structure.

3D surface structure estimation using neural networks for autonomous systems and applications

Growing community of inventors

Saratoga, CA, United States of America

Minwoo Park