Update README.md

README.md

@@ -47,14 +47,23 @@ catkin build
 conda activate cosypose
 cd ~/catkin_ws
 source devel/setup.bash
-roslaunch ros_cosypose singleview.launch
+roslaunch ros_cosypose singleview_rviz.launch
 ```
 * __Note:__ The camera must be calibrated for good results. See [this section](#usb-camera-calibration)
 
 ### Details
 
-#### object_pose_estimation.py node :
-This node subscribe to a rectified color image topic (default is `/usb_cam_0/image_rect_color`) and run cosypose on the latest image and publish the results on `/cosyobject_list` topic. (then loop).
+#### pose_estimation.py node :
+This node creates a service `/pose_estimation`. It expects, as an input for each call :
+* The camera intrinsic matrix
+* The image
+* **Optionally** A list of `CosyObjects` for warm-starting the pose estimation.  
+(If this list is not empty, then the detection phased is skipped and the pose estimation is only run on the given objects.)
+
+#### pose_estimation_loop.py node :
+This node simply calls the `/pose_estimation` indefinitely.  
+It subscribes to a camera image topic, and run the pose estimation in loop on the latest image. It then publishes the results on the `/cosyobject_list` topic.  
+(It can also "track" objects by warm-starting every estimation with the result of the previous one)
 
 #### rviz_translator.py node :
 This node subscribe to the `/cosyobject_list` topic (published by object_pose_estimation.py node) and transforms the incoming messages to other format that are displayable by rviz. By default, it publishes on :

scripts/pose_estimation_loop.py

@@ -32,7 +32,7 @@ class PoseEstimationLoop:
                 rospy.sleep(1.0)
 
 if __name__ == '__main__':
-    rospy.init_node('object_pose_estimation', anonymous=True)
+    rospy.init_node('pose_estimation_loop', anonymous=True)
     camera_name = rospy.get_param('~camera_name', "usb_cam_0")
     allow_tracking = rospy.get_param('~allow_tracking', "false")
     PoseEstimationLoop(camera_name, allow_tracking).main()