Handwritten Digit Recognition by csavas
Recognizes handwritten digits. Very fast, about 2.5 seconds at 60 fps. Draw into the WALL (thick light-gray) box using INWR (electronics section), anything else will break it. The start button can be clicked WITHOUT switching to sprk.
neural
network
intelligence
number
artificial
electronics
handwriting
subframe
mnist
Comments
-
_killer_ 24th Aug 2024
R33sesK1ng: From what I can tell, no, that would make it ten times harder. It uses a databank storing the weights for sections and evaluates the similarity and thus probability for each digit (the piston chambers) and outputs the most probable one. A very simple AI, incapable of learning. But the fact that this was made in a sandbox game is incredible. -
R33sesK1ng 24th Aug 2024
Wait does this use neural learning? -
mawi 24th Aug 2024
This is MAD impressive. Even when coding this in a normal programming language this is extremely impressive. The fact that you did it in Powder toy makes it 10x more impressive. Well done! -
KnZahid300 24th Aug 2024
idk how it is even possible but congratulations! -
blueskies1818 24th Aug 2024
AI....................... +3000 -
Arkymarble 24th Aug 2024
WPAAAAAAAAAH -
R33sesK1ng 23rd Aug 2024+1 for very sick +Fav for very cool
-
TheSingularity 23rd Aug 2024
Deserves 2nd page, +1 -
MultipleParticle 23rd Aug 2024
_killer_: i would rather like this on 2nd page -
Chipboard 23rd Aug 2024
Seriously impressive, wow.