Houdini 20.5 Nodes Geometry nodes

Labs ML CV Label Metadata 1.0 geometry node

Prepares the label metadata.

On this page
Since 20.5

This node prepares the label metadata that will be written to a COCO JSON at render time. It also prepares the ID attribute for Karma to render the appropriate ground truth data.

Optionally this node takes the output of the Labs ML CV Keypoint Metadata in the second input, which prepares the mesh/keypoints for the Labs ML CV ROP Annotation Output SOP and associates the keypoints with the label created by this node.

The original ML Computer Vision tools were developed by the Synthetic Data team at Endava PLC.

Parameters

Group

A subset of points in the input geometry to run the program on. Leave this blank to affect all points in the input.

Group Type

What the group is made of (point, vertex, primitive etc.)

Category ID

Category ID

The integer that will represent the category in the COCO JSON.

Category Name

The string name of the category that will be associated with the integer category ID.

Enable Instance ID

Enables the creation of the Instance ID attribute.

Instance ID

ID Type

The type of instance ID to create.

ID from Index

Instance ID will match the Category ID, generally used for single objects.

ID from Attribute

Instance ID will be generated from an attribute in Houdini such as crowd ID or connectivity, used when multiple objects exist for the same category.

Instance ID

An integer used to distinguish objects within the same category as separate entities. This value is the pixel value rendered in the ID pass from the Karma GT ROP and used to generate COCO bounding boxes and segmentation.

Examples

Tip

When viewing in Houdini’s Help Browser, please copy the example file’s URL to a regular browser to proceed with the download.

See also

Geometry nodes