Hi all,

as you probably know by now, OpenCV3 got released a few months ago. To
summarize what matters to the ROS community, OpenCV3 is:
- a big cleanup of the C API (which also comes with forward compatibility
brought to OpenCV 2.4.9+; meaning it's usually easy to write code that
compiles with OpenCV 2 and 3, like Python did between 2.7 and 3.x)
- more performance (many algorithms optimized, transparent GPU utilization
...)
- new algorithms especially in opencv_apps, cf the GSOC videos since 2013:
( [1], [2], [3] ), especially a few that seem very relevant to the ROS
community: RGBD processing, 3d visualization, deep learning support, object
recognition, calibration (faster, omni-cam support ...), better/faster
stereo ....
And more details in the announcement [4]

Now, all that seems like a very good reason to switch to OpenCV3 by default
for the next release of ROS: Kinetic. Please, let's discuss it on the SIG
as it might get technical:
https://groups.google.com/forum/#!topic/ros-sig-perception/K5__71SX7eU
(and feel free to create non-technical threads about the subject on the SIG
too)


[1] https://www.youtube.com/watch?v=_TTtN4frMEA
[2] https://www.youtube.com/watch?v=3f76HCHJJRA
[3] https://www.youtube.com/watch?v=OUbUFn71S4s
[4] http://opencv.org/opencv-3-0.html