Configuration Elements
A configuration consists of elements of three categories, shown as the three columns of the graph overview. Every element has a unique ID, its own settings in the editor and its own preview. The connections between the elements define the data flow of the application: camera → object detection → robot.
Cameras
A camera element stands for one physical camera - usually a LumiScanX - and the way it acquires
data: identification of the camera, exposure, depth estimation and the corrections applied to the
camera data. A calibration file (.hdvcal) belongs to each camera; upload it with Upload and
select it in the editor.
- Preview
-
The live image of the camera, and the point cloud computed from it, both with the current settings. Use it to check exposure, the depth range and the corrections before any detection is configured.
- Exported as
-
the
camerasandrecordingssections of the LumiScan VGR settings - see Camera Settings in the LumiScan VGR manual. - Connections
-
A camera is the source of one or more object detections.
Several cameras can be configured when the application uses more than one, for example one camera
per container (cam_main, cam_bags).
Object detections
An object detection element describes how objects are found in the data of the connected camera: which neural network is used, which objects are to be detected and localised, and how - in 2D, in 3D or with both combined - and how the results are filtered and sorted.
- Preview
-
The selected detection with the current settings on the current camera data.
- Exported as
-
the
networks,objectsandpipelinessections of the LumiScan VGR settings - see Network Settings, Object Settings and Pipeline Settings in the LumiScan VGR manual. - Connections
-
Input from one camera; output to one or more robots.
Neural networks are trained with the LumiScan Training Framework on images annotated with the LumiScan Annotation Framework.
Robots
A robot element describes a robot - or one robot with one particular gripper - that acts on the
results of the connected object detection: the robot description (.urdf), the hand-eye
calibration that places the robot relative to the camera, the gripper, and the positions and
motion segments used to approach and execute a grip. If collision avoidance is used, the 3D data
of robot, gripper and container are configured here as well.
- Preview
-
The robot in relation to the camera data with the current settings.
- Exported as
-
the
robotssection of the LumiScan VGR settings - see Robot Settings in the LumiScan VGR manual. - Connections
-
Input from one or more object detections.
A physical robot can be represented by several robot elements, for example one per gripper.