optical flow in opencv_apps

[ros-perception]

optical flow in opencv_apps

[fujimo-t]

I created small ROS packages for DIS Flow using OpenCV implementation and consider to create pull request to opencv_apps.

But FlowArray.msg is not suitable for dense flow as pointed in ros-perception/vision_opencv#67 (comment) .
Is it OK to add new msg for dense flow if I send pull request?

My current implementation is just wrapping sensor_msgs/Image like stereo_msgs/DisparityImage.
The reason of this is that Image has enough information to handle data array, and can be accessed easily by cv_bridge.

[k-okada]

Patches are always welcome, so feel free to create new PR.

• I prefer not to add new messages, especially just for one node, because we have to create viewer, test code and so on just for that node
• I like the comparison of algorithms, so if your new node outputs the same message as other existing nodes, then we're easy to compare the outputs
  • so if you can implement multiple dense optical flow algorithm, that's would be good
  • or enable to output new message for all existing optical flow algorithm

[fujimo-t]

I see.

Then I will create PR using existing message opencv_apps/Flow.msg.
It is not convinience for dense flow but also can be used.

Later I try to implement other dense algorithms or change to new message if I have time.