Hello!
I am using the carter_sim application and trying to generate training data for my robot’s object recognition, adding a class from static meshes was pretty easy, as I just added the class in the json config file, following this documentation: https://docs.nvidia.com/isaac/isaac_sim/plugins/isaac_sim/sensors.html?highlight=detection#enabling-bounding-boxes.
However, when I try to do the same for a spawned robot it is not that simple, I spawned a carter robot, named ‘carter_2’, from URDF by adding it to the apps/carter_sim/bridge_config/carter_full_config.json file, I changed the camera config to look as follows:
"segmentation_classes": {
"static_mesh": [
...
{
"name": "table"
},
{
"name": "carter_2"
}
]
},
"bounding_box_settings": {
"all_bounding_boxes": false,
"occlusion_check": true,
"output_occlusion_ratio": true,
"ray_spacing_in_pixels": 10,
"max_rays_per_box": 16,
"occlusion_threshold": 0.8,
"draw_debug": false,
"classes": [
...
{
"name": "table"
},
{
"name": "carter_2"
}
],
"exclude_classes": [
{
"name": "dummycharacter"
}
]
}
...
When I check on the Issac Sight app, the segmentation seems to work properly, showing the robot’s shade on a different color, but on the detection window, there is no bounding box for the carter robot, as you can see on the attached image.