- Jul 31, 2018
- 4,195
- 3,699
- 136
For the technically inclined you can roll your own .
There is plenty of C++ examples out there that connects to an IP camera and consumes the stream, with a little help from ffmpeg/libav its pretty straight forward to decode the stream realtime in code too, from here its only your imagination that sets the boundries, realtime motion detection etc and where you wanna store your videos.
I've done mine so I have motion detection and some other metadata that I store realtime, and then I am able to re-run detection/AI/plugin I've made on the metadata OR the real video itself over and over again. This comes in handy if you got action in one part of the feed, you put that part in a "detection box" (almost like enhance heh) and rerun with different plugin settings (higher sens, ai, whatever, this means you dont get false negatives on the rest of the frame) .. plus now I can run it 14 days back and see how many times the neighbors cat has used this path.
On my todo list is to encrypt the vids and store them offsite, right now end storage is just a local ftp server.
Listen to me and listen to me carefully. Fuck the cloud.
If you gonna cloud anyway, I would crack the software and reroute the remote storage destination to something local, encrypt, then move to alternate off site cloud.
Of course its a matter of time before the AI part of this thing is going decentralized and only work on your cloud data.. and then you're fucked.
I'm trying to figure out how to create an interface, that will allow me to view my cameras on my PC, rather than my phone or tablet. For one thing, the screen's a lot bigger. I'm sure there's got to be some kind of Android to Windows conversion tool box, I just haven't found it.
Yet.