Vision processing


Any suggestions on a package to handle vision on both the bebop and mambo in a general purpose manner? Python 3 & Opencv are fighting. My mamboVision package only works on python 2 and it drops a LOT of frames (which seems to be opencv’s fault in processing them). VLC has no issues processing the vision but it is a stand-alone program and I want the frames brought into python so my library can use them.

@Djavan or @Nicolas help?



Sorry I don’t really understand what you’re asking… :confused:


Sorry, I realize that question was far too inside my code to make sense. Let me ask it differently. Do you know of general purpose vision libraries that will process the vision streams in both of the formats that your drones use? So far as I can tell, Bebop is RTP and Mambo is RTSP. You have specialized code that handles the vision stream and displays it to the phones or iPads, right? I am having trouble with opencv handling the protocols right so I’m trying to figure out what other general purpose vision software exists.



I don’t know any libraries that can do that, sorry.

But for information, format of the video stream is quite the same: it is h264 encoded.
Both stream are received through the RTP protocol.
However, the RTP session start protocol is not the same: Bebop is started with ARSDK commands (and port are given during the json exchange, before the SDK connection), Mambo is started with RTSP.


Thanks. We have an idea of how to process a different way and are trying to figure out how to make it work for both drones. I’ll post with progress :slight_smile: