Change webcam resolution using v4l2

[rekumar]

I have mounted a USB endoscope to the gantry and am able to capture images using the following command:

ffmpeg -y -f video4linux2 -i /dev/video2 --ss 0:0:1 -frames 1 /data/user_storage/image.jpg

Despite being a 5 megapixel camera that registers properly on my other machines, the images I'm capturing with this command are at a 160x120 resolution. I believe I need to change the format using v4l2-ctl --set-fmt-video={width}x{height}, but it looks like v4l2 is not found in this version of Linux. Adding the -s 1024x768 command in ffmpeg as per the Opentrons docs stretches the image without changing the capture resolution.

Is there a way, v4l2 or otherwise, to change the format of the webcam stream within the opentrons version of linux?

PS -- I found a group from BNL that reported using v4l2 to control the format of a Logitech webcam mounted to their OT-2. Could this be a recent change to opentrons' buildroot? For context we just received our robot in early Jan 2024.

Thank you and let me know if any more info could help!

[sfoster1]

Huh. We haven't really touched anything about the webcam stuff in quite a while, and I don't think we've ever had v4l2-ctl built and installed.

You could try using this python library: https://pypi.org/project/v4l2py . Since v4l2 is fundamentally a kernel driver API, you interact with it through ioctls, and python has a really good story for that, so this library should be installable on an OT-2 and doesn't actually contain any C code and in fact should be callable from a protocol.

This will be a little more annoying to use, but it looks like what you'd do (and unfortunately this is just from the docs for me - v4l2 devices are different enough that you'll have to see what your endoscope reports) is something like this, in a python shell (including a Jupyter notebook shell on the OT-2):

>>> from v4l2py.device import Device, BufferType
>>> cam = Device.from_id(2)  # This should be the same number as /dev/videoN
>>> cam.open()
>>> print(cam.info.card)  # this should probably print the model to confirm it's the right one
>>> cam.set_format(BufferType.VIDEO_CAPTURE, width=1024, height=768)

And that should do it.