Surface estimation
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 parametric mathematical modeling. A variety of synthetic 3D road surfaces may be generated by modeling a 3D road surface using varied parameters to simulate changes in road direction and lateral surface slope. In an example embodiment, a synthetic 3D road surface may be created by modeling a longitudinal 3D curve and expanding the longitudinal 3D curve to a 3D surface, and the resulting synthetic 3D surface may be sampled to form a synthetic ground truth projection image (e.g., a 2D height map). To generate corresponding input training data, a known pattern that represents which pixels may remain unobserved during 3D structure estimation may be generated and applied to a ground truth projection image to simulate a corresponding sparse projection image.
1 . One or more processors comprising processing circuitry to:
generate a first representation of a synthetic three-dimensional (3D) road surface of a synthetic road based at least on sampling a lateral 3D curve for two or more points on a longitudinal 3D curve representing the synthetic road; and
augment a dataset with the first representation of the synthetic 3D road surface.
2 . The one or more processors of claim 1 , wherein the processing circuitry is further to compute one or more lateral values of the longitudinal 3D curve based at least on simulating one or more changes in road direction of the synthetic road.
3 . The one or more processors of claim 1 , wherein the processing circuitry is further to compute one or more height values of the longitudinal 3D curve based at least on simulating one or more changes to a height at one or more points of the synthetic 3D road surface.
4 . The one or more processors of claim 1 , wherein the processing circuitry is further to compute one or more lateral values of the longitudinal 3D curve as a second order polynomial of one or more longitudinal values of the longitudinal 3D curve using one or more values for one or more polynomial constants sampled to simulate changes in road direction of the synthetic road.
5 . The one or more processors of claim 1 , wherein the processing circuitry is further to simulate one or more changes to a height at one or more points of the synthetic 3D road surface based at least on computing one or more height values of the longitudinal 3D curve as a linear combination of Fourier bases, using one or more sampled values for at least one of: a number of the Fourier bases, a weight for a particular basis of the Fourier bases, or a frequency for the particular basis.
6 . The one or more processors of claim 1 , wherein the processing circuitry is further to generate the lateral 3D curve for each point of at least one of the two or more points on the longitudinal 3D curve based at least on sampling a value for angle between the lateral 3D curve and a ground plane to simulate one or more changes in lateral surface slope of the synthetic road.
7 . The one or more processors of claim 1 , wherein the processing circuitry is further to generate an incomplete representation of the synthetic 3D road surface based at least on setting a subset of pixel values of the first representation of the synthetic 3D road surface to zero to simulate one or more unobserved values, and associate the incomplete representation with the first representation of the synthetic 3D road surface in the dataset.
8 . The one or more processors of claim 1 , wherein the processing circuitry is further to:
generate a plurality of binary maps representing a pattern of unobserved values associated with estimating 3D surface structure from real-world data; and
generate an incomplete representation of the synthetic 3D road surface based at least on multiplying the first representation of the synthetic 3D road surface by at least one of the binary maps.
9 . The one or more processors of claim 1 , wherein the processing circuitry is further to use one or more neural networks trained using the dataset to perform at least one of:
adapting a suspension system of an ego-object,
navigating the ego-object to avoid a detected protuberance, or
applying an acceleration or deceleration to the ego-object based at least on a detected surface slope.
10 . The one or more processors of claim 1 , wherein the one or more processors are comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
11 . A system comprising one or more processors to:
generate a ground truth representation of a synthetic three-dimensional (3D) road surface of a synthetic road based at least on sampling a lateral 3D curve for two or more points on a longitudinal 3D curve representing the synthetic road; and
update a dataset to include the ground truth representation of the synthetic 3D road surface.
12 . The system of claim 11 , wherein the one or more processors are further to compute one or more lateral values of the longitudinal 3D curve based at least on simulating one or more changes in road direction of the synthetic road.
13 . The system of claim 11 , wherein the one or more processors are further to compute one or more height values of the longitudinal 3D curve based at least on simulating changes to a height at one or more points of the synthetic 3D road surface.
14 . The system of claim 11 , wherein the one or more processors are further to compute one or more lateral values of the longitudinal 3D curve as a second order polynomial of one or more longitudinal values of the longitudinal 3D curve using one or more values for one or more polynomial constants sampled to simulate changes in road direction of the synthetic road.
15 . The system of claim 11 , wherein the one or more processors are further to simulate one or more changes to a height at one or more points of the synthetic 3D road surface based at least on computing one or more height values of the longitudinal 3D curve as a linear combination of Fourier bases, using one or more sampled values for at least one of: a number of the Fourier bases, a weight for a particular basis of the Fourier bases, or a frequency for the particular basis.
16 . The system of claim 11 , wherein the one or more processors are further to generate the lateral 3D curve for each point of at least one of the two or more points on the longitudinal 3D curve based at least on sampling a value for angle between the lateral 3D curve and a ground plane to simulate one or more changes in lateral surface slope of the synthetic road.
17 . The system of claim 11 , wherein the one or more processors are further to generate an incomplete representation of the synthetic 3D road surface based at least on setting a subset of pixel values of the ground truth representation of the synthetic 3D road surface to zero to simulate one or more unobserved values, and associate the incomplete representation with the ground truth representation of the synthetic 3D road surface in the dataset.
18 . The system of claim 11 , wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
19 . A method comprising:
generating, using one or more processors, a ground truth representation of a synthetic three-dimensional (3D) road surface of a synthetic road based at least on sampling a lateral 3D curve for two or more points on a longitudinal 3D curve representing the synthetic road; and
associating, using the one or more processors, the ground truth representation of the synthetic 3D road surface with a dataset.
20 . The method of claim 19 , wherein the method is performed by at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.