#Neuroomba
1 messages · Page 1 of 1 (latest)
race the roomba? https://aws.amazon.com/deepracer/league/
A roomba with a Go-pro is certainly an option but its probably better to go for some sort of RC car with suspension and larger tires. Roombas arent known for having very good driving characteristics outside of perfectly flat rooms.
Actually, an RC Tank would probably work the best, if you ask me. :V
Somebody.. quick... 3d print a Neuro funko to ride the tank
I wonder if you could 3D print that VR chat model someone made
yes, Unity export to FBX, import it to Blender then export to STL
I may have to do that...
What aboyt using supersonic sensors instead of a camera?
I mean, i assumed the idea was to have Nero roll around and explore the world, she cant do that without eyes.
She doesn't have eyes?
Supersonic Sensors arent the same as a camera, its more of a radar. You cant see your surroundings with it, so much as detect the presence of it.
i like this one, its doesn't use a path painted on floor https://www.youtube.com/watch?v=u7oDqWJhXR0
Github repo: https://github.com/mmajewsk/Tonic
Demo video 1: https://youtu.be/W2pxHu_dPf8
Demo video 2: https://youtu.be/FkcIxE0a1w8
Big Thanks to Hackerspace Kraków: https://hackerspace-krk.pl/
Music: www.bensound.com
Pedestrain crossing footage from: www.videvo.net
uses very little processing, and is very easy to program
Its great as one of the tech in place
Yeah but could her image recognition software use it as a source?
Like, she looks at a turtle and says "Oh hi Vedal"
Or does the supersonic sensor only tell her "That is the shape of a turtle"
Well its absolutely fantastic for depth perception
Tbe entire point of an AI is to learn to correspond certain things to different data
And having a 2d map of ypur environment as you walk around isn't such a bad idea now is it
I could see it being a good idea on top of having cameras, maybe
But for the sake of streaming entertainment its just a non-starter to have Neuro roam around without Chat being able to see what she's looking at
Well of course having a camera is crucial
But sensors rrplace what a camera won't give you
Aka depth perveption
Depends on the cameras, but yeah fair enough- I've also seen robots using ChatGPT prompts to figure out movement patterns for themselves, so its definitely possible, but we'd need someone willing to try it
There are transformer models that can take images as input, though for them to be good they require tremendous processing power and are mostly closed source
Google has PaLM (I think it's the one? not sure), OpenAI has GPT-4 though they haven't rolled out the image as input feature for the public yet.