I'm attempting to stream the 8MP and/or VGA cameras using GStreamer, but typical pipelines don't work as I get a
Device '/dev/video0' cannot capture at 1280x720
or any other resolution I choose. I understand there is a sample capture application, but I need to know the caps that will work for each camera in order to use my application.
Alternatively, has anyone been able to use either camera with OpenCV's VideoCapture() class?
Thank you for interest in the Intel® Aero Compute Board.
Could you please tell me what guide you have been using?
Also, what is the OS of your host machine?
I'll be waiting for your response.
The only guide I've been using have been the meta-intel-aero Wiki on Github (where I found the gstreamer pipeline for /dev/video13) as well as this forum. However I have not yet seen any documentation on what gstreamer pipeline I can use on /dev/video0 to stream the VGA or 8MP cameras (the capture_example code only grabs frames from those cameras, as far as I can tell, it does not stream them, which is what I require).
I'm using Ubuntu 14.04 LTS on my host machine.
I second that. It would be great if someone at Intel provides us with a primitive working gstreamer pipeline for the 8MP and VGA camera. Similar to this pipeline for Realsense:
gst-launch-1.0 v4l2src device=/dev/video13 ! xvimagesink
Thank you for the information provided.
I understand that you want documentation related to what GStreamer pipeline to use on /dev/video0 to stream the cameras.
I'll investigate if that information is available, and if it isn't, I'll inform the corresponding team of your request.
Either way, I'll post my findings here.
Thank you for your patience.
Any updates? I'm very much looking forward to evaluating the Aero board but I cannot test my software without access to the VGA/8MP cameras via GStreamer or OpenCV.
We don't have an estimate time of arrival for the example code, I am currently investigating but I don't have any updates yet. As soon as I find something relevant I'll post it here.
Sorry for the inconvenience this may cause.
Thank you for your patience.
Maybe this can help you. https://github.com/intel-aero/sample-apps/pull/1 Optical Flow (OpenCV) using downward facing VGA camera via V4L2 by broody · Pull Request # 1 · intel-aero/sample-apps · G…
It grabs raw frames from the download facing VGA camera via V4L2 and then feeds it to OpenCV to do optical flow.
Thanks Broody, i'm giving this a try. My concern though is that there is quite a bit of initialization and configuration just to pull frames from the device. How can this be ported to work with OpenCV's VideoCapture class? Alternatively, could you provide a gstreamer pipeline that we could use with the VideoCapture class?
VideoCapture class also uses V4L2 calls. You should be able to modify the initialization that's done there (https://github.com/opencv/opencv/blob/master/modules/videoio/src/cap_libv4l.cpp opencv/cap_libv4l.cpp at master · opencv/opencv · GitHub ) and add in the additional V4L2 calls required for Aero (which can be found in init_device() of the sample app).
Thats helpful, thank you. A couple of follow up questions:
1.) The color on the 8MP camera looks off, or at least the exposure is very low. Is it possible to enable auto-exposure?
2.) Any thoughts on a Gstreamer pipeline/source element?
1) Will have to defer to Intel support team on this. I do see the same low exposure issue however.
2) I'm not very experienced with gstreamer but looking at source, the same V4L2 modifications may be needed. https://github.com/GStreamer/gst-plugins-good/blob/master/sys/v4l2/gstv4l2object.c gst-plugins-good/gstv4l2object.c at master · GStreamer/gst-plugins-good · GitHub If you run "gst-inspect-1.0 v4l2src" there is also a prepare-format option which seem to serve this purpose.