A Wave Function Collapse Overlapping Model Guide • Hacker News Board

Re: A Wave Function Collapse Overlapping Model Guide

#11

Post by 42yeah1 » Fri, Jan 31, 2020, 10:17 AM UTC

symmetricsaurus wrote: post #10

Seems like n=3 really helps to get good results. With n=2 it was a bit hard to see that it actually followed the pattern.

Yeah, that's why in the official repo lots of images were generated using N=3. However N=2 is pretty good for mazes and stuffs from my perspective.

Re: A Wave Function Collapse Overlapping Model Guide

#12

Post by BorisTheBrave » Fri, Jan 31, 2020, 10:38 AM UTC

Note that the implementation described above uses the Arc Consistency 3 algorithm for the propagate step. It's often an order of magnitude faster to use AC4, as is done in the original WFC repo and in my implementation (https://github.com/BorisTheBrave/DeBroglie).

Re: A Wave Function Collapse Overlapping Model Guide

#13

Post by 42yeah1 » Fri, Jan 31, 2020, 5:33 PM UTC

BorisTheBrave wrote: post #12

Note that the implementation described above uses the Arc Consistency 3 algorithm for the propagate step. It's often an order of magnitude faster to use AC4, as is done in the original WFC repo and in my implementation ( https://github.com/BorisTheBrave/DeBroglie ).

Yeah, I am just trying to make it work :P I would strongly advise against using my implementation.

Re: A Wave Function Collapse Overlapping Model Guide

#14

Post by skolskoly » Fri, Jan 31, 2020, 6:01 PM UTC

42yeah1 wrote: post #9

skolskoly wrote: post #7

Earlier quoted context omitted.

I broke it in a bad way. Pattern chugged along fine and then it froze and a random white rectangle appeared. (N=3) https://imgur.com/a/IlmdkC6

Ah, sorry I didn't make that clear... That would be a contradiction, as no pattern was found to fit in that area. You need to generate again.

Interesting. Well it only happened once so I guess I just got bad luck. Fun demo!