How to get the 3D Bounding Box of the detected object using Deep Learning?

edit

Hi

I would like to perform collision checking/obstacle detection using stereo camera. Im using the darknet_ros package link text. So , would like to do some simulation for collision checking in Moveit. Firstly, would like include some objects in the scene and need the 3D bounding Box of the detected objects using Deep Learning CNN such as YOLO or Retinanet. Im using ROS kinetic and Ubuntu 16.04. Any help how to get the 3D Bounding box using this package?

Thanks