Dev/image encoding param #10
Conversation
Authors: - Marko Durkovic <marko@ternaris.com> - Florian Friesdorf <florian@ternaris.com>
Allows other OpenCV image encodings
This provides at least a bit of a clue regarding the configuration of nodes, complementing the already existing examples for the various "detail" views
marv_robotics/nodes/camera_frames.py
Outdated
| @@ -59,7 +60,7 @@ def camera_frames(bagmeta, messages, image_width=320, max_frames=50): | |||
| continue | |||
|
|
|||
| try: | |||
| img = imgmsg_to_cv2(msg, "rgb8") | |||
| img = imgmsg_to_cv2(msg, encoding) | |||
There was a problem hiding this comment.
If I'm not mistaken, then the argument you specify here is the "target" encoding, why do you want to change that?
The encoding that the ROS image has is already part of the Image msg and should be already considered by the conversion function....
There was a problem hiding this comment.
Hmm, now that you mention it, that makes sense. That said, the existing code, when it deals with our bag files, generates incorrect images (blueish grass etc.) Adding this code and switching to "bgr8" fixed the problem. Though now I'm uncertain why.
There was a problem hiding this comment.
I would double check that your data is indeed encoded bgr8 and that the encoding string is also set correctly. Otherwise, can you provide a sample bag file?
There was a problem hiding this comment.
All video images in my Marv instance show a R/B and B/R confusion as well. Using bgr8 as target encoding makes sense to me. OpenCV uses BGR encoding internally and methods like cv2.imwrite want BGR encoded images.
There was a problem hiding this comment.
@Earthwings, do you use the latest release? If so, I would check which code path is taken here https://github.com/ternaris/marv-robotics/blob/master/marv_robotics/cam.py#L124-L130 and open a new issue in case the bags look fine e.g. in image_view or rviz.
There was a problem hiding this comment.
@NikolausDemmel Works indeed fine in 17.11.
Unfortunately I'm running into multiple other bugs with that release, but that's off topic here.
This branch provides the ability to specify the OpenCV encoding for the processing of of Image topics (ours are in bgr8 format). Also includes a node configuration example in the marv.conf.in template.