voxl-streamer client repetitive connect/disconnect

Eric Katzfey

An update to UVC camera support has been started. It will increase number of supported cameras by allowing MJPEG and will increase number of supported viewers including latest QGroundControl versions. Target availability is mid-December.

wilkinsaf

@Eric-Katzfey Thank you for the update Eric!

Also, is there a format/guide on how to modify the camera configuration file? For example if a camera was giving off a different format, framerate, size, etc...

I was trying to find a guide on how to modify that file, but could not.

I figured it would be helpful for developers.

Eric Katzfey

@wilkinsaf The configuration file for voxl-streamer is documented here: https://docs.modalai.com/voxl-streamer/#configuration-file-format

Eric Katzfey

@wilkinsaf The configuration for voxl-camera-server is discussed here: https://docs.modalai.com/voxl-camera-server/#configuration

Eric Katzfey

@wilkinsaf There is an initial version of a new UVC camera server application now available from our "dev" package repo. The application voxl-uvc-server works in conjunction with voxl-streamer to better support UVC cameras. It has been tested with the Logitech C270 and Videology 5MP Micro USB 2.0 Board Camera. The documentation is at https://docs.modalai.com/voxl-uvc-server/ and the code is at https://gitlab.com/voxl-public/modal-pipe-architecture/voxl-uvc-server. To try it out, install the latest Voxl platform release voxl_platform_3-3-0-0.5.0-a. On target, configure the opkg package manager to use the modalai dev repo (https://docs.modalai.com/configure-opkg/). Then use opkg install voxl-streamer to upgrade voxl-streamer from 0.2.6 to 0.2.7 (and libmodal_pipe from 2.1.1 to 2.1.2). Then opkg install voxl-uvc-server to install the new voxl-uvc-server application v0.0.5. Run voxl-uvc-server -d to start the application with debug messages. Then run voxl-streamer with voxl-streamer -f /usr/share/modalai/voxl-streamer/voxl-streamer.conf -c uvc-server to use the new configuration file and record type for the uvc server. Let me know if this works for you.

wilkinsaf

quick comment for anyone reading this, it is

voxl-uvc-server -d when running

Eric Katzfey

@wilkinsaf Thanks! I fixed the typo in the original post.

wilkinsaf

@Eric-Katzfey

I am seeing that this failed for MJPEG

voxl:~$ voxl-uvc-server -d
Enabling debug messages
voxl-uvc-server starting
UVC initialized
Device found
Device opened
uvc_get_stream_ctrl_format_size failed for MJPEG
uvc_get_stream_ctrl_format_size failed

Is there a config file for this that I can modify to change the format?

Eric Katzfey

@wilkinsaf That shows a failure of the camera to support either MJPEG or YUV422. In order to support that camera you need to know what formats it supports. The best way to do that is to query it's capabilities on the command line. First of all use the lsusb command to find out the device id and product id of the camera. For example:

Ubuntu18$ lsusb
Bus 001 Device 090: ID 046d:0825 Logitech, Inc. Webcam C270
Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub

This shows you that the UVC webcam has device id and product id 046d:0825. You can use that to query it's capabilities with lsusb -v -d 046d:0825. This will dump all of the camera capabilities to the output. Can you respond with the output from your camera?

wilkinsaf

@Eric-Katzfey

Apologies for the delay. Some other things came up.
I was able to pull down the camera information. We have 2 cameras we are trying to get up.

Cam 1: https://pastebin.com/qHvHKXay
Cam 2: https://pastebin.com/7DgvP58k

They appear to be able to post in MJPEG

wilkinsaf

Not sure if voxl-uvc-server got removed.
I can not seem to find it in the public repo:

Eric Katzfey

@wilkinsaf Project is here: https://gitlab.com/voxl-public/modal-pipe-architecture/voxl-uvc-server

Eric Katzfey

@wilkinsaf The current voxl-uvc-server capabilities only cover uncompressed frames, not mjpeg. The two cameras that you are looking at both have uncompressed formats but the guidFormat is slightly different than the cameras I have tested with. Let me look into this a little deeper.

Eric Katzfey

@wilkinsaf The formats should be compatible. I'll have to dig into the libuvc code to see why they are not accepted.

Eric Katzfey

@wilkinsaf Are either of these cameras cheap and easy to purchase?

wilkinsaf

I believe so. Let me find them online really quick. I was able to get the video's from the cameras streaming using the directions here: https://docs.modalai.com/uvc-streaming/

I moved over the ffmpeg ipk and executed
ffmpeg -i /dev/video2 -f mpegts udp://QCS_IP:4242

So I know that they are working.

Eric Katzfey

@wilkinsaf There are some clues here: https://github.com/gregkh/usbutils/issues/42. It seems like there are some endian issues with interpretation of the guidFormat field. For example, for my Logitech C270 test camera, lsusb reports the guidFormat as 32595559-0000-0010-8000-00aa00389b71 but libuvc (which voxl-uvc-server uses) reports the guidFormat as 59555932-0000-1000-8000-00aa00389b71. So, my guess is that since lsusb reports the guidFormat of your camera as 59555932-0000-1000-8000-00aa00389b71, libuvc (and hence voxl-uvc-server) will see it as 32595559-0000-0010-8000-00aa00389b71 which doesn't match any valid formats. I think I can hack libuvc to add the extra guidFormat and then your cameras will work but I won't be able to test it without one of your cameras. But I will post it and you can try it. For reference this is where libuvc needs to be modified to add the extra guidFormat: https://github.com/libuvc/libuvc/blob/37f00abd32b083a3d748f2a25c1c794578b51c16/src/stream.c#L109

Eric Katzfey

@wilkinsaf There is a new version of libuvc available at http://voxl-packages.modalai.com/dev/libuvc_1.0.2_202112182223.ipk with the changes. If you rebuild the dev branch of voxl-uvc-server and install build deps from dev you should pick up the new libuvc. Then you can rebuild and test with your camera.

wilkinsaf

@Eric-Katzfey roger that. thank you for your support here Eric

wilkinsaf

@Eric-Katzfey will post results after run