search

ros-perception / vision_msgs

Algorithm-agnostic computer vision message types for ROS.
Apache License 2.0
155 stars 74 forks source link

unifying 2D and 3D messages #2

Closed procopiostein closed 7 years ago

procopiostein commented 7 years ago

IMO I think we gain in simplicity by reducing the number of messages by merging, when possible, 2D and 3D msgs. A pose can be interpreted as 2D or 3D according to the value on the z field. For example, in many cases, wheeled robots in ROS publish their pose as a geometry_msgs/Pose, (inherently 3D) although they are always at the ground level.

Kukanani commented 7 years ago

+1 on combining ObjectHypothesis. The DetectionXD messages will also have to be updated to point to the new message type.

Thanks for the PR!

procopiostein commented 7 years ago

Agreed @tfoote , I have not thought about that. I reverted the related commit and tried to improved the description and fields. Also changed the datatype for the size of a 2D BB to uint32, as it should be in pixels.

@Kukanani I updated the DetectionXD to use the generic ObjectHypothesisWithPose

mintar commented 7 years ago

Let's continue the discussion in #5.