dlCHOMPDataOptions

Options for generating training and validation data samples

Since R2024a

Description

The dlCHOMPDataOptions object enables you to specify options for generating training and validation samples using the generateSamples function.

The dlCHOMPDataOptions object requires the Deep Learning Toolbox™.

Creation

Syntax

opts = dlCHOMPDataOptions

opts = dlCHOMPDataOptions(Name=Value)

Description

opts = dlCHOMPDataOptions returns default options for generating training and validation samples.

opts = dlCHOMPDataOptions(Name=Value) returns default options for generating training and validation samples.

Properties

RadiusRangeOfObstacles — Minimum and maximum radius for spherical obstacles in sample generation environment

[0.2 0.5] (default) | two-element vector of positive scalars

Radius range for the spherical obstacles in the sample generation environment, specified as a two-element vector of positive scalars, in the form [MinimumRadius MaximumRadius]. The maximum radius must be greater than the minimum radius value.

Set RadiusRangeOfObstacles based on the size of obstacles in the robot workspace.

Each of the obstacles in each of the data sample JSON files that generateSamples generates during the environment-creation phase has a random chance of being in the specified radius range based on the RandomStream and Seed property values.

CountRangeOfObstacles — Minimum and maximum number of spherical obstacles in the sample generation environment

[20 30] (default) | two-element vector of positive integers

Minimum and maximum number of spherical obstacles in the sample generation environment, specified as a two-element vector of positive scalars, in the form [MinNumSpheres MinNumSpheres]. The maximum number of spherical obstacles must be greater than the minimum number of spherical obstacles.

Set CountRangeOfObstacles based on the number of and size of obstacles in the robot workspace.

The number of obstacles in each of the data sample JSON files that generateSamples generates during the environment-creation phase has a random chance of being in the specified range based on the RandomStream and Seed property values.

MinDistanceFromBase — Minimum distance that spherical obstacles must be from robot base in sample generation environment

0.2 (default) | positive numeric scalar

Minimum distance that spherical obstacles must be from the robot base in the sample generation environment, specified as a positive numeric scalar. The distance is measured from the origin of the robot base to the surface mesh of an obstacle, in meters.

NumSamples — Number of desired samples to generate

2000 (default) | positive integer

Number of desired samples to generate, specified as a positive integer.

ValidationSplit — Percentage of training samples to include in validation dataset

0.2 (default) | numeric scalar in range [0, 1]

Percentage of training samples to put in validation dataset, specified as a numeric scalar in the range [0, 1].

RandomStream — Random number algorithm

"threefry" (default) | "global"

Random number algorithm, specified as either "threefry" or "global".

• "threefry" — Use the Threefry 4x64 generator with 20 rounds for sample generation.

• "global" — Use the global random number stream for sample generation. You can get and set the global number stream using the RandStream.getGlobalStream and RandStream.setGlobalStream functions, respectively. When RandomStream is set to "global", dlCHOMPDataOptions does not use the Seed property.

For more information on generator algorithms, see Creating and Controlling a Random Number Stream.

Seed — Random number seed

0 (default) | nonnegative integer

Random number seed, specified as a nonnegative integer.

Version History

Introduced in R2024a