Openset 3D detection pipeline gripping point strategy #681
Description
Is your feature request related to a problem? Please describe.
Currently, the 3D detection system calculates the centroid of detected objects, assuming the gripper doesn't need to rotate and that the fingers are wide enough for a grip. However, for certain use cases, such as with vacuum grippers, we need to focus solely on the top plane of the object (e.g., a
box). This functionality is currently unavailable.
Describe the solution you'd like
Robust (various camera frames) ransac based filtering
Describe alternatives you've considered
Other algos
Additional context
Add any other context or screenshots about the feature request here.