Beginning....
I am aiming to write what I am calling "a toy" (the word game is too heavy for me at the moment). This toy will be a computer graphics application which involves launching pods that grow plant like things (i.e. lots of opportunities for me to play with geometry).
A few caveats:
-all assets (i.e. items my program is currently drawing) are stand-ins (perhaps some art students will want to build me some nice models at some point)
-yes, I intend the final one to be in 3D
-I am human
After thinking about it too much, I finally started with implementing Poisson sampling (in 2D in Processing), because I have to start somewhere.
I have started with R. Bridson's 'Fast Poisson Disk Sampling in Arbitrary Dimensions', because he was kind when I met him once and shared his course notes on animation and it is a nice short paper (and as an academic, I seem to like digesting papers more then tutorials - even though this meant spending overly too much time thinking about this sentence "Generate up to k points chosen uniformly from the spherical annulus between radius r and 2r around xi", because if the theta angle sampling is uniform, is the radius varying between r and 2r also uniformly distributed for the k samples or random, I ended up with uniform).
I fell into all the typical implementing challenges as I went ahead and tried to write the whole thing in one go (i.e. without unit tests).
First I had a bug that resulted in not adding samples near the extents of the space.
Second, just as I was wrestling with myself about going back to write unit tests, my computer ran out of battery and restarted and I lost several days of work (as I am working on this in very tiny momentary breaks so progress was slow and will be very slow).
After wasting time being very mad with myself for not having a github set up (there is one now), I sat down and re-wrote more slowly with more testing along the way. I am now happy to have relatively pretty 2D sampling. (Keen is of course right on this topic).
But when I ask my husband which image he likes better, random or Poisson, he picked random... hmmm.... well at least the Poisson gives me a better 'growth' look over time.
What do you think? (videos below, random first, then Poisson)
A few caveats:
-all assets (i.e. items my program is currently drawing) are stand-ins (perhaps some art students will want to build me some nice models at some point)
-yes, I intend the final one to be in 3D
-I am human
After thinking about it too much, I finally started with implementing Poisson sampling (in 2D in Processing), because I have to start somewhere.
I have started with R. Bridson's 'Fast Poisson Disk Sampling in Arbitrary Dimensions', because he was kind when I met him once and shared his course notes on animation and it is a nice short paper (and as an academic, I seem to like digesting papers more then tutorials - even though this meant spending overly too much time thinking about this sentence "Generate up to k points chosen uniformly from the spherical annulus between radius r and 2r around xi", because if the theta angle sampling is uniform, is the radius varying between r and 2r also uniformly distributed for the k samples or random, I ended up with uniform).
I fell into all the typical implementing challenges as I went ahead and tried to write the whole thing in one go (i.e. without unit tests).
First I had a bug that resulted in not adding samples near the extents of the space.
Why aren't their samples near the edges? |
Second, just as I was wrestling with myself about going back to write unit tests, my computer ran out of battery and restarted and I lost several days of work (as I am working on this in very tiny momentary breaks so progress was slow and will be very slow).
After wasting time being very mad with myself for not having a github set up (there is one now), I sat down and re-wrote more slowly with more testing along the way. I am now happy to have relatively pretty 2D sampling. (Keen is of course right on this topic).
Pretty samples |
But when I ask my husband which image he likes better, random or Poisson, he picked random... hmmm.... well at least the Poisson gives me a better 'growth' look over time.
Random sampling | Poisson Sampling |
Comments
Post a Comment