Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?
This is your opportunity to ask questions. No question too simple or too silly.
Culture war topics are accepted, and proposals for a better intro post are appreciated.
Jump in the discussion.
No email address required.
Notes -
How hard is it to put a neural network-based targeting system onto a small UAV? Yes, yes, Watchbird by Sheckley and all that, but the radio link with the drone operator is the weakest point of modern UAVs.
In Australia we just expelled a Chinese student for working on linking drones together to navigate through environments without GPS, since it was apparently 'WMD-related'. You should tread carefully, a lot of people are getting very excited/scared about drone navigation and targeting.
Honestly they're not wrong, I'm waiting for the moment 10,000 kill bots swarm out of a shipping container somewhere, they'll make 9/11 look like a joke.
More options
Context Copy link
[caveat: there are some parts of this discussion area I'm not going to describe, either publicly or in private messages.]
Any? Throw YOLO onto a Jetson, some batteries, and a decent lens, have fun. Start with one of a dozen preconfigured models, maybe use the VisDrone dataset if you want to put a bit of effort into tweaking something. Probably about 2k-10k if you're buying all pre-built (maybe a bit more if you want non-DJI).
Actually doing it well gets harder. There's been a little bit of effort into image-based localization, to avoid GPS jamming, but for technical reasons (off-nadir sunlight suuuuuucks) it's annoying as hell to work with, and you probably want it or something like it, in addition to your main sensor payload. Some human identification is easier than others. And actual decision-making for more complex stuff is a Coding Project.
More options
Context Copy link
More options
Context Copy link