Might not look like much, but I trained an RBM I wrote to automatically generate Pokemon: I also wrote a convolutional layer, but I'm not done debugging it. When that's finished the output should be cleaner. Surprisingly, a mean filter didn't do much to improve quality.
|
|
# ? Aug 23, 2015 04:40 |
|
|
# ? May 9, 2024 07:41 |
|
Jo posted:Surprisingly, a mean filter didn't do much to improve quality. Maybe you should try a nice filter instead.
|
# ? Aug 23, 2015 04:42 |
|
Avenging Dentist posted:Maybe you should try a nice filter instead. Or, more helpfully, a median filter.
|
# ? Aug 23, 2015 07:13 |
Hubis posted:Or, more helpfully, a median filter. Much obliged. I'll give this a try. I'm still trying to grok the difference in their behavior.
|
|
# ? Aug 23, 2015 17:57 |
|
Jo posted:Might not look like much, but I trained an RBM I wrote to automatically generate Pokemon: the thing on the left is nightmarish. pass it through a hell filter to enhance the blood coming out of its mouth's eyes. good work!!!
|
# ? Aug 23, 2015 21:21 |
|
Jo posted:Much obliged. I'll give this a try. I'm still trying to grok the difference in their behavior. this book is a really good into to implementing filters, and apparently free now! http://www.dspguide.com
|
# ? Aug 23, 2015 21:24 |
|
Jo posted:Much obliged. I'll give this a try. I'm still trying to grok the difference in their behavior. The median is the 'middle value' of the set (i.e. the one in the middle when all values are sorted based on some criteria) wheras the (arithmatic) mean is the average of the set. Since the median value will match one of the adjacent pixels, there will be no interpolation/blurring, so you'll preserve sharp edges in the data set. This is good if you're upsampling something from a limited palette and want the output to stay in that palette, for example. Of course you can only really do this if you have a reasonable way to sort the set to find the median in the first place (and that can be the tricky part for a median filter) but in your case it would be an easy implementation since the image is already black-and-white.
|
# ? Aug 24, 2015 11:17 |
|
I am writing my master's thesis in LaTeX and I need to typeset both Haskell and Scala code inline in the document, literate style. Normally Haskell is handled by lhs2tex, but it's extremely old and kind of lovely, and is a bitch to integrate into your main document if you only need a few pages typeset. So I spent two days porting the useful functionality of lhs2tex (which is only a very small part of it, turns out) to Python and integrating it with pygments to make it handle any language rather than just Haskell. It also simply processes text so you can process two paragraphs out of your whole document, rather than producing a full separate TeX document. The desirability of non-monospaced typesetting of code may be controversial but I thought it was an interesting pursuit anyway and I'm pretty happy with the result. The important feature is the column-based alignment from lhs2tex. It's not as fine-tuned as lhs2tex but I'm just going to tweak it ad-hoc to fit my needs as I write my thesis. The Scala part mostly works but I haven't written a token spacing algorithm for it yet so everything is all mushed togther (other than the column-based alignment). I dunno why that object keyword is not highlighted, pygments knows about it---I checked. Volte fucked around with this message at 13:37 on Aug 24, 2015 |
# ? Aug 24, 2015 13:30 |
|
Well that's satisfying. Still have to tweak some of the symbol replacements and spacing and stuff, but now I can get back to writing.
|
# ? Aug 24, 2015 15:56 |
|
I'm back.
|
# ? Aug 25, 2015 18:31 |
|
clockwork automaton posted:I'm back. that is some akira poo poo right there. please tell me this wasn't intended.
|
# ? Aug 25, 2015 19:30 |
|
clockwork automaton posted:I'm back.
|
# ? Aug 26, 2015 15:32 |
|
Jo posted:Might not look like much, but I trained an RBM I wrote to automatically generate Pokemon: If it doesn't work, just remember, Missingno was a legit Pokemon.
|
# ? Aug 26, 2015 17:18 |
clockwork automaton posted:I'm back. That is some unsettling metaphysical poo poo right there.
|
|
# ? Aug 26, 2015 19:49 |
movax posted:If it doesn't work, just remember, Missingno was a legit Pokemon. Actually, I should check and make sure the data set doesn't have Missingno in it. That might be causing problems.
|
|
# ? Aug 26, 2015 20:15 |
|
Internet Janitor posted:A while back I posted about a K interpreter I've been writing. Last night I started tinkering with a project I call iKe: This is cool! Yeah the language is ridiculously expressive. I've been trying to learn K and KDB for a while now. There's some documentation of course but I found most to be quite dense and difficult to follow. Are there any learning resources you could recommend?
|
# ? Aug 27, 2015 14:29 |
|
mobby_6kl: It is a bit difficult to find good resources. I learned most of what I know about the language by poring over the K2 Reference Manual as I wrote my K interpreter. The Wiki for Kona, the other open-source K interpreter, has links to a ton of articles and other material. No Stinking Loops is another great place to find interesting K programs, although some are presented without much discussion and thus rather cryptic. For my own part, I have written a brief but reasonably comprehensive manual for oK which describes the verb and adverb behaviors I have reverse-engineered by studying k5. I also wrote a bit about K programming techniques. iKe is coming along nicely since my last posts. A guy who works for Dyalog wrote an asteroids game, and I have been tinkering with an implementation of breakout: In the Pixel Art thread I wrote a quick one-off image processing program to experiment with animation effects: And I also whipped up one way to implement conway's game of Life:
|
# ? Aug 27, 2015 15:40 |
|
Rendering Sierpinski's Triangle using the Chaos Game method. I find the initial random point inside the triangle's three points by getting Barycentric coordinates. using Javascript lol
|
# ? Aug 27, 2015 16:49 |
|
LP0 ON FIRE posted:Rendering Sierpinski's Triangle using the Chaos Game method. I find the initial random point inside the triangle's three points by getting Barycentric coordinates. Chaos Game bro. Here's my Chaos Game shits: http://forums.somethingawful.com/showthread.php?threadid=3676849#post441073931 e: Hopefully I'll have some new Chaos Game progress in September, since I'm about finished with some of my (non-screenshot-worthy) projects. Avenging Dentist fucked around with this message at 19:16 on Aug 27, 2015 |
# ? Aug 27, 2015 18:49 |
|
That's really neat! I had a go at writing one in K: http://johnearnest.github.io/ok/ike/ike.html?gist=8c2ec75f1fe34c4cd898 FIrst I define a function which calculates random iterative successors: code:
code:
Internet Janitor fucked around with this message at 19:23 on Aug 27, 2015 |
# ? Aug 27, 2015 19:14 |
|
Avenging Dentist posted:Chaos Game bro. Here's my Chaos Game shits: http://forums.somethingawful.com/showthread.php?threadid=3676849#post441073931 Those are actually amazing looking! Are those Lorenz Attractors too? And holy crap, the post above that. (Though totally different kind of fractal.) Internet Janitor posted:That's really neat! I had a go at writing one in K: Really nice. What kind of results do you get if you use round instead of floor values? e: Never mind I guess I can just edit that code, and it's sort of blowing my mind how little code there is. LP0 ON FIRE fucked around with this message at 19:53 on Aug 27, 2015 |
# ? Aug 27, 2015 19:45 |
|
I imagine it would look pretty much the same- I only floor at the end of my calculations. iKe is designed for working with relatively low resolution bitmaps. 160x160 pixels is all it can draw at for the time being. edit: iKe's manual and more example code is here if anyone wants to tinker. Internet Janitor fucked around with this message at 20:00 on Aug 27, 2015 |
# ? Aug 27, 2015 19:50 |
|
Internet Janitor posted:I imagine it would look pretty much the same- I only floor at the end of my calculations. iKe is designed for working with relatively low resolution bitmaps. 160x160 pixels is all it can draw at for the time being. Pretty interesting! I often turn to Javascript to tinker or prototype, but this sounds like a good thing to use as well. By the way, I tried round if I did it correctly: _ 0.5 + whatever I made a nicer looking one than what I had before using low alpha pixels and rendered it during my lunch break: Now I want to make a huge one overnight and print it 300dpi. My life is exciting.
|
# ? Aug 27, 2015 20:11 |
|
LP0 ON FIRE posted:Those are actually amazing looking! Are those Lorenz Attractors too? And holy crap, the post above that. (Though totally different kind of fractal.) Not specifically Lorenz, but obviously they are attractors. I'm using this paper as a basis for my project.
|
# ? Aug 27, 2015 23:03 |
|
Wrote a Twitter Bot for parsing the Leper's Colony. It scans it every 5 minutes or so, runs a query using another library I wrote to see if there are new entries and, if so, posts them to Twitter. I'm trying to think of how to get the reason and actual post in there too. I can't fit the reason into a tweet, unless I post it elsewhere. So I was thinking of just making a screenshot of the post and adding the reason on top of it. Not sure if I can render HTML in a console Web Job, but it's probably worth a shot.
|
# ? Aug 28, 2015 15:09 |
|
Drastic Actions posted:
Have you looked at phantomjs? Basically a headless, scriptable browser that you can also use to capture screenshots, among other things. (http://phantomjs.org/screen-capture.html) Edit: Looks like there's a C# wrapper as well: http://www.nrecosite.com/phantomjs_wrapper_net.aspx
|
# ? Aug 28, 2015 18:57 |
|
Kumquat posted:Have you looked at phantomjs? Basically a headless, scriptable browser that you can also use to capture screenshots, among other things. (http://phantomjs.org/screen-capture.html) Their ImageGenerator library works... kinda. The forums have pretty hosed up markup, so their generator only gets some of the css properly, so it won't get the full post. I made a makeshift post generator of my own (a weaker version of what I have on my forums app) and it worked better. However the bigger problem is that it won't work on Azure. The image never finishes processing . In their FAQ it says it should work, so I'll have to ask them. But thanks for sending me in the right direction
|
# ? Aug 28, 2015 21:30 |
|
gently caress you, boxes! Ha!
|
# ? Aug 31, 2015 22:13 |
Crosspost from Making Games thread. So I hate how paint.net has like no sprite or animation support in it. So I hated it so much, that I made a thing. It loads .pdn files, and lets you turn on/off layers and specify animation stuff. It also watches the file you open, so you can make a quick tweak in paint.net, and it will automatically reload it and start animating the new changes.
|
|
# ? Aug 31, 2015 23:51 |
lord funk posted:gently caress you, boxes! Ha! This is cool. How many objects, and how fast is it running? A while back, I was fiddling with image manipulation. These are old, and I've developed better filters, but these are the screenshots I have handy. Most of these are just two-dimensional plus/minus matrices applied to the image. I don't know what I'm doing, but it's fun. Edit: Orzo posted:A bunch of random recent screenshots from my game: Dude, whatever happened to this? It was looking so kickass. Centripetal Horse fucked around with this message at 01:22 on Sep 2, 2015 |
|
# ? Sep 2, 2015 01:12 |
|
Centripetal Horse posted:
Yeah, I really wish there would be some updates on this too. This is all hearsay, but apparently he had a kid, and it took over his spare time.
|
# ? Sep 2, 2015 16:52 |
|
Centripetal Horse posted:This is cool. How many objects, and how fast is it running? So far up to 6000 boxes at 60fps. Pretty basic lighting (ambient + diffused + specular), and the biggest bottleneck I'm getting is poor array performance in Swift for the box models.
|
# ? Sep 2, 2015 18:31 |
|
lord funk posted:So far up to 6000 boxes at 60fps. Pretty basic lighting (ambient + diffused + specular), and the biggest bottleneck I'm getting is poor array performance in Swift for the box models. Are you using metal?
|
# ? Sep 2, 2015 18:39 |
|
Tres Burritos posted:Are you using metal? Yeah. So far really liking it, and the debugging tools are pretty excellent.
|
# ? Sep 2, 2015 18:54 |
LP0 ON FIRE posted:Yeah, I really wish there would be some updates on this too. His gain is our loss, I guess. That project looked great. I hope it hasn't just died on the vine. lord funk posted:So far up to 6000 boxes at 60fps. Pretty basic lighting (ambient + diffused + specular), and the biggest bottleneck I'm getting is poor array performance in Swift for the box models. Is that running on a mobile device? Which part of the array performance is slowing you down? Are you adding and removing a lot of objects at arbitrary locations?
|
|
# ? Sep 2, 2015 20:40 |
|
Centripetal Horse posted:Is that running on a mobile device? Which part of the array performance is slowing you down? Are you adding and removing a lot of objects at arbitrary locations? Yeah, iPad. I'm seeing excessive retain / release of my models while I iterate through them to update the physics. If I were using Obj-C, I'd turn off ARC to solve this. But I'm not sure what to do in Swift land.
|
# ? Sep 2, 2015 22:23 |
|
LP0 ON FIRE posted:This is all hearsay, but apparently he had a kid, and it took over his spare time.
|
# ? Sep 3, 2015 09:56 |
|
After reading The C++ Programming Language 4th edition and watching a bunch of 8 out of 10 Cats does Countdown, I decided to code a couple of cheat programs to check out the new C++11 features. The words part was pretty easy. Take a dictionary, copy_if every word that can be formed with the given letters (using a map<char,int> to keep track), and sort by length. code:
The huge list of possibilities was then sorted by accuracy, numbers used and operations used to create the most accurate result, using as few numbers, and as simple operators as possible. A simple print function evaluates the best result again, converting the RPN to a more familiar form. code:
|
# ? Sep 3, 2015 12:18 |
|
Programmer Humor posted:After reading The C++ Programming Language 4th edition and watching a bunch of 8 out of 10 Cats does Countdown, I decided to code a couple of cheat programs to check out the new C++11 features. What? Oh its a tv game show. I get it now. That actually sounds like something that would be a question in a google code jam.
|
# ? Sep 3, 2015 17:32 |
|
|
# ? May 9, 2024 07:41 |
|
Zaphod42 posted:What? 8 out of 10 cats is a comedy show, the actual show is called Countdown. [link]https://www.youtube.com/watch?v=_JQYYz92-Uk[/link]
|
# ? Sep 4, 2015 09:42 |