ROS2 and Fusion API

ml22 · May 23, 2024, 1:10pm

Is it currently possible to create a ROS2 node that does not directly attach to a camera but subscribes to camera data from other nodes and uses the fusion API to perform multi camera body tracking on that data?

If so, what topics would it make sense to subscribe to?

Myzhar · May 23, 2024, 1:31pm

Hi @ml22
Welcome to the Stereolabs community.

Our roadmap currently does not include porting the Fusion module to the ROS Wrapper 2, only partially supporting GNSS in a single-camera setup.

You could create your ROS 2 wrapper around the sl::Fusion class that retrieves the information from “N” running ROS 2 Nodes.
You must also modify the code of the sl::ZedCamera component to enable the publishing of the camera information to the Fusion module:

github.com

stereolabs/zed-sdk/blob/master/object detection/multi-camera/cpp/src/ClientPublisher.cpp#L78


      
          
          
              return true;
          }
          
          void ClientPublisher::start()
          {
              if (zed.isOpened()) {
                  running = true;
                  // the camera should stream its data so the fusion can subscibe to it to gather the detected body and others metadata needed for the process.
                  zed.startPublishing();
                  // the thread can start to process the camera grab in background
                  runner = std::thread(&ClientPublisher::work, this);
              }
          }
          
          void ClientPublisher::stop()
          {
              running = false;
              if (runner.joinable())
                  runner.join();