Distance estimation to objects and free-space boundaries in autonomous machine applications
In various examples, a deep neural network (DNN) is trained—using image data alone—to accurately predict distances to objects, obstacles, and/or a detected free-space boundary. The DNN may be trained with ground truth data that is generated using sensor data representative of motion of an ego-vehicle and/or sensor data from any number of depth predicting sensors—such as, without limitation, RADAR sensors, LIDAR sensors, and/or SONAR sensors. The DNN may be trained using two or more loss functions each corresponding to a particular portion of the environment that depth is predicted for, such that—in deployment—more accurate depth estimates for objects, obstacles, and/or the detected free-space boundary are computed by the DNN. In some embodiments, a sampling algorithm may be used to sample depth values corresponding to an input resolution of the DNN from a predicted depth map of the DNN at an output resolution of the DNN.
1. A processor comprising:
processing circuitry to:
compute, using a deep neural network (DNN) and based at least in part on sensor data generated using one or more sensors of an ego-machine, first data representative of first depth values corresponding to one or more objects and second data representative of second depth values corresponding to one or more free-space boundaries; and
perform one or more operations for controlling the ego-machine based at least in part on the first depth values and the second depth values.
2. The processor of claim 1 , further comprising processing circuitry to:
compute one or more locations of one or more bounding shapes corresponding to the one more objects; and
associate the first depth values with the one or more bounding shapes based at least in part on the one or more locations,
wherein the performance of the one or more operations is further based at least in part on the association of the first depth values with the one or more bounding shapes.
3. The processor of claim 2 , wherein the computation of the one or more locations is executed using the DNN, another DNN, or a computer vision algorithm.
4. The processor of claim 1 , further comprising processing circuitry to:
compute one or more locations of the one or more free-space boundaries; and
associate the second depth values with the one or more free-space boundaries based at least in part on the one or more locations,
wherein the performance of the one or more operations is further based at least part on the association of the second depth values with the one or more locations of the one or more free-space boundaries.
5. The processor of claim 4 , wherein the computation of the one or more locations is executed using the DNN, another DNN, or a computer vision algorithm.
6. The processor of claim 1 , wherein the one or more operations include at least one of path planning, world model management, obstacle or collision avoidance, a control decision, or an advanced driver assistance system (ADAS) operation.
7. The processor of claim 1 , wherein the DNN is trained using a first loss function corresponding to depth values of free-space boundaries and a second loss function corresponding to depth values for object detections.
8. The processor of claim 1 , wherein the sensor data corresponds to image data generated using one more image sensors of the ego-machine, and the DNN is trained using training data generated using one or more depth sensors.
9. The processor of claim 1 , wherein the computation, using the DNN, includes computing data representative of one or more depth maps representative of the first depth values and the second depth values.
10. The processor of claim 1 , wherein the processor 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 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 sensors;
one or more memory units; and
one or more processing units comprising processing circuitry to:
compute, using a deep neural network (DNN) and based at least in part on sensor data generated using the one or more sensors, data representative of one or more depth maps indicative of depth values;
associate a first set of the depth values with one or more free-space boundaries and a second set of the depth values with one or more detected objects; and
determining one or more control operations based at least in part on the association.
12. The system of claim 11 , further comprising processing circuitry to:
compute one or more locations of one or more bounding shapes corresponding to the one more detected objects,
wherein the association of the first set of the depth values with the one or more detected objects is based at least in part on associating the first set of the depth values the one or more locations.
13. The system of claim 12 , wherein the computation of the one or more locations is executed using the DNN, another DNN, or a computer vision algorithm.
14. The system of claim 11 , further comprising processing circuitry to:
compute one or more locations of the one or more free-space boundaries; and
wherein the association of the second set of the depth values with the one or more free-space boundaries is based at least in part on associating the second set of the depth values the one or more locations.
15. The system of claim 14 , wherein the computation of the one or more locations is executed using the DNN, another DNN, or a computer vision algorithm.
16. The system of claim 11 , wherein the sensor data corresponds to image data generated using one more image sensors of the one or more sensors, and the DNN is trained using training data generated using one or more depth sensors.
17. 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 implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
18. A processor comprising:
processing circuitry to cause performance of one more control operations of an ego-machine based at least in part on depth values associated with one or more free-space boundaries, a deep neural network (DNN) generating data indicating that the depth values are associated with the one or more free-space boundaries.
19. The processor of claim 1 , wherein the second data further includes an indication that the second depth values correspond to the one or more free-space boundaries.
20. The processor of claim 1 , wherein the one or more free-space boundaries are associated with an area of an environment for which the ego-machine is able to navigate without colliding with the one or more objects.