|
Started easing back into gamedev after a hiatus. I'm happy with my progress on this shmup/twinstick thing so far ![]() ![]() A "high quality" version: http://i.imgur.com/G9fnCsH.gif
|
![]() |
|
![]()
|
# ? Jun 17, 2024 22:22 |
|
xezton posted:Started easing back into gamedev after a hiatus. I'm happy with my progress on this shmup/twinstick thing so far This is very nice. Reminds me of Team 17's Project X
|
![]() |
Continuing to work on my convolutional autoencoder. First step: overfit a small training example. Success!![]()
|
|
![]() |
|
watch in amazement as this random noise becomes... a meme!! *audience oohs*
|
![]() |
|
Decided to start tinkering with Vulkan yesterday by writing a raymarcher with deferred shading. Current state: it's basically a crappy shadertoy full of boilerplate but, hey, pixels on screen.![]()
|
![]() |
|
Suspicious Dish posted:watch in amazement as this random noise becomes... a meme!! *trains a deep-learning net on an endless supply of internet memes* *back-feeds a white noise signal into the outputs*
|
![]() |
|
the network dreams of memes
|
![]() |
|
Xerophyte posted:Decided to start tinkering with Vulkan yesterday by writing a raymarcher with deferred shading. Current state: it's basically a crappy shadertoy full of boilerplate but, hey, pixels on screen. This on github anywhere?
|
![]() |
Hubis posted:*trains a deep-learning net on an endless supply of internet memes* Please do not share the inner workings of my consciousness, tia.
|
|
![]() |
|
Baloogan posted:the network dreams of memes code:
|
![]() |
|
Tres Burritos posted:This on github anywhere? Nah. It's a slightly massaged vulkan tutorial with a shadertoy attached at this point, one that has about a 50% chance of crashing my display driver every time I run it. I added some better SDFs and materials along with shadows and AO today so I'm getting to the point where the renders are pretty fun, though: ![]() The quantization errors are annoyingly persistent..
|
![]() |
|
I've had a Minecraft/Zelda thing stuck in head for a while so I figured it would be a good excuse to finally sit down and start learning basic Unity stuff. Early prototype stages obviously. Not much to do but to walk/swim around the Perlin terrain for now, but I made a handful of biomes so it's kind of cool. :3![]()
|
![]() |
|
crab avatar posted:I've had a Minecraft/Zelda thing stuck in head for a while so I figured it would be a good excuse to finally sit down and start learning basic Unity stuff. Early prototype stages obviously. Not much to do but to walk/swim around the Perlin terrain for now, but I made a handful of biomes so it's kind of cool. :3 Kind of cool? No, that's cool as hell. Like really. Keep up on that.
|
![]() |
|
I'm training neural networks to play chess.![]() You can play against them online! The long term plan is to tun this into a full chess engine based on modern deep learning, but I'm still in the pretty early stages. Right now I have a network that looks at the board and directly predicts the best move to play. It doesn't do any search, which means sometimes it occasionally makes silly moves, but it already plays a lot better than I was expecting.
|
![]() |
|
By network do you mean a neural net? And by searching you mean you're not looking up future board states, but going purely off the neural net's judgement? Pretty cool stuff. I played with neural nets in college and it really blew my mind how easy they are to set up and how powerful they are. Especially with supervised learning you can really quickly reach a very good solution. What you could do is have like a trainer program which does traditional deep search, but do that after the neural net commits to an action, and then use that as supervised learning for faster reinforcement.
|
![]() |
|
Zaphod42 posted:By network do you mean a neural net? And by searching you mean you're not looking up future board states, but going purely off the neural net's judgement? Pretty cool stuff. Yes that's exactly what I mean. The plan for doing search is to use a neural network like I have now (right now it actually ranks all possible moves behind the scenes and chooses the best one to play) to choose which branches of the tree to search, and also to train a second network to predict the outcome of the game and use that as the evaluation function.
|
![]() |
|
Nippashish posted:Yes that's exactly what I mean. The plan for doing search is to use a neural network like I have now (right now it actually ranks all possible moves behind the scenes and chooses the best one to play) to choose which branches of the tree to search, and also to train a second network to predict the outcome of the game and use that as the evaluation function. Mmmm, nested neural networks. I love it.
|
![]() |
|
Nippashish posted:I'm training neural networks to play chess. Yeah this thing keeps kicking my rear end. I just can't beat it. It even adapts just fine to trying to throw it off with completely irrational plays.
|
![]() |
|
Working on this game about purple merfolk butts with a bunch of other guys right now:![]() Still very rough, but it's getting there. Core gameplay (5 vs 5 TRPG arena gladiator matches) is working fine and we're refocusing towards football manager -style team management.
|
![]() |
|
Working on shadow mapping:![]()
|
![]() |
|
So we're getting closer to reaching a 1.0 milestone with our D&D 5e apps. I've completed phone layouts for our character sheet app, d20: My original intent was a pure tablet experience because I wanted to translate the 3 page character sheets from the Handbook into digital form but a lot of our group uses phones or 7" tablets. I also reworked a lot of the Spellbook since I previously wasn't accounting for spellslots, and the (not pictured) view for adding spells lacked a lot of useful functionality like filtering by class or level. Last post I mentioned that we had a very early stage app for our Dungeon Master. It's less early stage now though still a little rough: We've totally gutted the device-to-device communication. It was originally an experiment with Google's Nearby APIs but during a 3-4 hour D&D session we found that Nearby would drain your battery like no other. We've since rewritten the networking layer to do very basic mDNS discovery and then connect via websocket. Players using d20 are automatically joined to sessions hosted by Master. The Android TV app (Tavern) I showed last time displays information from both apps. The DM is able to draw out the play field on his tablet, which is mirrored on the TV display, and he can manage initiative order as well. We plan on letting him add background images so that the canvas isn't an ugly white all the time. I spent some time today triaging bugs for a 1.0 milestone which will be when we release. We absolutely will opensource the Github repo, and possibly also push to the Play Store.
|
![]() |
|
HappyHippo posted:Working on shadow mapping:
|
![]() |
|
That all looks very cool!Ferg posted:We've since rewritten the networking layer to do very basic mDNS discovery and then connect via websocket. How hard was this to do? I've wondered about connecting Android and iOS via mDNS.
|
![]() |
|
pokeyman posted:That all looks very cool! Dead simple. We found this: http://ennova-it.github.io/ZeRXconf/
|
![]() |
|
Shalinor posted:Cascaded? or horizon mapping? Etc? Just the most basic kind (following this tutorial, but with WebGL). I might try Poisson Sampling or something to soften the edges.
|
![]() |
|
Ferg posted:Dead simple. We found this: http://ennova-it.github.io/ZeRXconf/ Perfect, thanks!
|
![]() |
|
![]() I picked up a side project this weekend and wrote GoonAuth2, an API server that can be used to verify that users are members of the SA forums. It's powered by Django Rest Framework and Redis, though I suffered from everything's-a-nail syndrome and initially used a regular database model with a management script and cronjob to make the hashes it generates temporary. Fortunately a couple of goons in the Django thread pointed me towards Redis and its ability to automatically expire records so I switched to using that instead. Of course in doing so I realized that I could replace the entire Django backend with a simple Flask app but oh well, I just love Rest Framework too much ![]() In addition to GoonAuth2 I also wrote a Discord bot for the SA Discord server that uses this API to automatically elevate users to a higher Role. When you first join the server you can't see anything but a waiting room. Before this weekend, you had to post your SA profile and tag a moderator, who then had to manually click the link to verify who you are. Now, the on-boarding process involves new users typing !authme <SA USERNAME> and getting PM'd by a bot with instructions on how to verify their identity: ![]() Once you pass the check, the bot adds the required Role that lets you see all of the other channels. It's not all that technologically impressive but I still learned a lot and can't believe I was able to pull it off over a weekend. I do plan on open-sourcing the GoonAuth2 codebase, but there's nothing stopping anyone from using the instance up and running on http://auth.everythingisgoon.com/. You can find instructions over there, so feel free to use it anywhere you might need to verify someone's gooniness ![]() IAmKale fucked around with this message at 05:48 on Mar 7, 2016 |
![]() |
|
Always happy to see more Forums scrapery! Nice job!
|
![]() |
Thanks in no small part to Nippashish's endless support in the academic thread, I was able to train a deep convolutional neural network to generate Pokemon. This is a big step up from the horrifying mess that I posted earlier. Here's a fun time lapse of the network's attempts to draw, starting from the earliest "just learning to recognize colors" into about 25000 iterations. It's past 500k now, so I might make another as it's farther along. http://i.imgur.com/gLDYNIK.webm
|
|
![]() |
|
Jo posted:Thanks in no small part to Nippashish's endless support in the academic thread, I was able to train a deep convolutional neural network to generate Pokemon. This is a big step up from the horrifying mess that I posted earlier. Here's a fun time lapse of the network's attempts to draw, starting from the earliest "just learning to recognize colors" into about 25000 iterations. It's past 500k now, so I might make another as it's farther along. It's Clefairy!
|
![]() |
|
I'm extremely excited to see how that goes! I've wanted to make a neural network for quite some time.
|
![]() |
|
Attempting to make a simple platformer... While I learn more about java/game programming..![]() ![]() ![]()
|
![]() |
|
Oh my god those boxes are so cute ![]()
|
![]() |
|
Cory Parsnipson posted:Oh my god those boxes are so cute Reminds me a little of the SuperCrateBox clone I started but never finished a few years ago ![]()
|
![]() |
|
cross-post from the 3D graphics thread...Hubis posted:If anyone is going to be at GDC this year, feel free to swing by and check out my talk Wednesday afternoon: Fast, Flexible, Physically-Based Volumetric Light Scattering
|
![]() |
|
Bob Morales posted:Reminds me a little of the SuperCrateBox clone I started but never finished a few years ago Those boxes are not as cute they barely even have eyeballs ![]() 3/5
|
![]() |
|
I've been watching some guy named Ferris work on an N64 emulator, and I decided to see if I could throw something together. I present to you: Rust8! ![]() Here's the github repo in case anyone's interested.
|
![]() |
|
Jo posted:Thanks in no small part to Nippashish's endless support in the academic thread, I was able to train a deep convolutional neural network to generate Pokemon. This is a big step up from the horrifying mess that I posted earlier. Here's a fun time lapse of the network's attempts to draw, starting from the earliest "just learning to recognize colors" into about 25000 iterations. It's past 500k now, so I might make another as it's farther along. I'd call it "drive by comments" rather than "endless support", but thanks ![]()
|
![]() |
|
nuvan posted:I've been watching some guy named Ferris work on an N64 emulator, and I decided to see if I could throw something together. Quick, to the Chip-8 Thread! Worth noting, if you used Cowgod's Chip-8 Technical Reference it describes the semantics of load/store and the shift instructions wrong. I recommend Mastering Chip8.
|
![]() |
|
![]()
|
# ? Jun 17, 2024 22:22 |
|
Finally got a software triangle rasterizer done. Doesn't do proper perspective texture mapping, simply affine texture mapping, but, well: https://zippy.gfycat.com/CompetentRepulsiveIberianemeraldlizard.webm Roughly 300loc, with boilerplate. http://magcius.github.io/sw3dv/ https://github.com/magcius/sw3dv
|
![]() |