|
Knotfest
|
# ? Jun 11, 2018 03:10 |
|
|
# ? Apr 24, 2024 23:50 |
|
Olive! posted:Knotfest Jesus Christ
|
# ? Jun 11, 2018 21:52 |
|
Neural network generated philosophy books: https://disexplications.tumblr.com/post/174769120729/i-took-another-look-at-the-openlibrary-data-dump-i Tag yourself, I'm What real world: a Phenomenology of managers
|
# ? Jun 13, 2018 00:08 |
|
I am Whitehead’s moral issues in nursing consciousness:quote:(...) Thus I establish that the value for the qualitative quality of existence in our universe is negative, ist est most living conscious entities suffer more than they enjoy living. Such subject has been noted multiple times not only in philosophy but also in art, as we are subject to the 'thousand natural shocks that flesh is heir to'. Elentor has a new favorite as of 01:58 on Jun 13, 2018 |
# ? Jun 13, 2018 00:33 |
|
Name of my old philosophy professor is on there lol. 2spooky
|
# ? Jun 13, 2018 13:00 |
|
SerialKilldeer posted:Tag yourself, I'm What real world: a Phenomenology of managers
|
# ? Jun 13, 2018 18:55 |
|
I'm torn between Existence of man, 1964-1977 and Essentialist approach to learning human human love personally.
|
# ? Jun 13, 2018 21:14 |
|
As a longtime lurker, I'm naturally On uncomments.
|
# ? Jun 13, 2018 21:55 |
|
DeepArt.io should really be a bigger thing in here. It's so easy to use and I'm sure you comedians on here could do it to great effect. Crossposting this from the GBS cat thread but note that your input photos do not actually have to be of cats!Dumb Lowtax posted:Everybody post DeepArt of your cats! You can use https://deepart.io/ on your cat pictures. The site mixes your cat photo with an art style, where an "art style" is just another arbitrary photo you upload.
|
# ? Jun 14, 2018 05:19 |
|
I tried mixing my cat with the crystal pepsi logo But mixing her with a scene from Space Dandy worked much better Or another cat with a vaporwave sunset Or my face, with an old stereo receiver faceplate or a crashing wave made of pugs Queen_Combat has a new favorite as of 06:19 on Jun 14, 2018 |
# ? Jun 14, 2018 06:14 |
|
Me in the style of van Gogh's self portrait: Same photo but using this image from the video game Cognition as the style. My cousin, van Gogh style: Tiggum has a new favorite as of 06:24 on Jun 14, 2018 |
# ? Jun 14, 2018 06:21 |
|
This space dandy style image is hot fire
|
# ? Jun 14, 2018 06:24 |
DeepArt has seemed really cool for a long time but I have to admit that it became a thing at a time when internet was really really buzzing about DeepDream and whatnot, so I never quite found it & wasn't sure if the pics I was seeing were just someone's personal NN projects or actually accessible tools. Thanks for linking it!
|
|
# ? Jun 14, 2018 14:11 |
|
https://twitter.com/KeatonPatti/status/1006961202998726665 https://twitter.com/christinelove/status/1002562010796699648 Also, I made a fractal crow:
|
# ? Jun 14, 2018 14:32 |
|
Dumb Lowtax posted:DeepArt.io should really be a bigger thing in here. It's so easy to use and I'm sure you comedians on here could do it to great effect. Crossposting this from the GBS cat thread but note that your input photos do not actually have to be of cats! It's not a cat, but I like how this turned out.
|
# ? Jun 14, 2018 21:28 |
|
Metal Geir Skogul posted:I tried mixing my cat with the crystal pepsi logo Metal Geir Skogul posted:This space dandy style image is hot fire These turned out super good, I think you should cross post them here now that the contest has ended and make the cat contest thread be about DeepArt
|
# ? Jun 14, 2018 21:37 |
|
|
# ? Jun 14, 2018 21:47 |
|
SerialKilldeer posted:Also, I made a fractal crow:
|
# ? Jun 14, 2018 22:18 |
|
Man I'd love to use deepart professionally if a single 2048x2048 image didn't cost 70 dollars.
|
# ? Jun 14, 2018 22:23 |
|
Elentor posted:Man I'd love to use deepart professionally if a single 2048x2048 image didn't cost 70 dollars. Jesus Christ, yeah, sell it for what it's worth (a little CPU time) not whatever you think you can extract out of the few most desperate customers. I'll bet they'd make more selling them for $5 a pop just by the sheer number of casual people who'd want one.
|
# ? Jun 14, 2018 22:29 |
|
all the code is open source for doing this, which means you just have to spend a good twenty or so hours getting everything configured, then you can make UNLIMITED seventy dollar images, walla!
|
# ? Jun 14, 2018 23:17 |
|
Watching a bunch of recent computerphile videos about neural networks and "deepart"-style stuff, they say that even on the most optimized code so far, downscaling the images, with the 6 or so class-identifiers that are required for the "art" part of it to work out (finding edges, objects, backgrounds, foregrounds, animals/humans, pupils, mouths), it chugs their 8x12GB graphics card machine to a standstill for a few minutes. They speculate that places like deepart get away with doing things in batches on drastically scaled hardware that is shared with a more profit-focused venture. Remember: doubling the dimensions quadruples the pixels and octuples edge-finding algorithms. Scale all that back to, say, a 2x8GB SLI'd gaming-hardware focused machine on non-server CPUs, and you'd be up poo poo creek with an $8 power bill for half a day.
|
# ? Jun 15, 2018 02:40 |
|
Metal Geir Skogul posted:Watching a bunch of recent computerphile videos about neural networks and "deepart"-style stuff, they say that even on the most optimized code so far, downscaling the images, with the 6 or so class-identifiers that are required for the "art" part of it to work out (finding edges, objects, backgrounds, foregrounds, animals/humans, pupils, mouths), it chugs their 8x12GB graphics card machine to a standstill for a few minutes. They speculate that places like deepart get away with doing things in batches on drastically scaled hardware that is shared with a more profit-focused venture. Remember: doubling the dimensions quadruples the pixels and octuples edge-finding algorithms. You can dramatically decrease processing requirements if you optimize your network. Like, by three orders of magnitude - you can do images in a couple seconds on a phone. https://jeffxtang.github.io/cnn,/deep/learning,/ai,/ios,/tensorflow/2017/02/19/fast-neural-style-on-ios.html
|
# ? Jun 15, 2018 04:46 |
|
|
# ? Jun 15, 2018 06:00 |
|
|
# ? Jun 16, 2018 01:59 |
|
I did some of that photo of petulant Trump at the G7 summit.
|
# ? Jun 16, 2018 18:47 |
|
|
# ? Jun 16, 2018 22:04 |
https://twitter.com/genekogan/status/1008338409025409024
|
|
# ? Jun 17, 2018 14:45 |
|
Finally caught up with this thread, and it owns.
|
# ? Jun 17, 2018 19:00 |
|
Procedurally generated representation and renderings of scenes https://www.youtube.com/watch?v=G-kWNQJ4idw https://deepmind.com/blog/neural-scene-representation-and-rendering/
|
# ? Jun 17, 2018 20:48 |
|
Hispanic! At The Disco posted:I assumed it was Mervyn Peake's fishing memoir. Nice. I appreciate this post. Just so you don't think you're posting into the void.
|
# ? Jun 17, 2018 21:24 |
|
Is there any work on real time Neural Network Anti-aliasing? That would be nice.
|
# ? Jun 17, 2018 23:32 |
|
Yeah, AI de-noising is a thing https://www.youtube.com/watch?v=YjjTPV2pXY0 https://www.youtube.com/watch?v=tjf-1BxpR9c
|
# ? Jun 18, 2018 04:10 |
Some more book titles! It feels like adding more names to the source database (can it be a corpus even though this is markov chain -based?) makes interesting titles more rare; one reason is probably that the more titles you have to generate from, the more "boring" options you have for the generation. Another is possibly that more titles also mean more different words in the database in general, so it's easier for longer words to become nonsense mix-and-match word salad. I quite like how the generator sometimes combines names and words, I've had "Leatherine", "Deatherine" and "Psychiatricia", at least.
|
|
# ? Jun 18, 2018 04:14 |
|
soundsection posted:Yeah, AI de-noising is a thing so cool I hope the stuff in that second video is coming to game engines soon, that comparison between SSAO and their solution is amazing
|
# ? Jun 18, 2018 05:59 |
|
ekuNNN posted:so cool If it's like the other gameworks features, it'll possibly be in Unreal Engine by the end of the year. Nvidia's been pretty good about keeping those ridiculous, high end features updated for UE, even if nearly nobody uses them due to how costly they are for the most part. https://www.youtube.com/watch?v=Ts76Y_828Hk
|
# ? Jun 18, 2018 07:51 |
|
Elentor posted:Is there any work on real time Neural Network Anti-aliasing? That would be nice. Nvidia has played with anti-aliasing traditional rasterizers using deep learning as well, which will probably be practical sooner than the raytracing/denoising stuff. http://openproblems.realtimerendering.com/s2017/05-msalvi-dl-future-of-rendering.pptx
|
# ? Jun 18, 2018 13:48 |
|
https://www.youtube.com/watch?v=MjViy6kyiqsquote:Researchers from NVIDIA developed a deep learning-based system that can produce high-quality slow-motion videos from a 30-frame-per-second video, outperforming various state-of-the-art methods that aim to do the same.
|
# ? Jun 18, 2018 17:40 |
|
That's drat impressive
|
# ? Jun 18, 2018 18:13 |
|
|
# ? Apr 24, 2024 23:50 |
|
repiv posted:Nvidia has played with anti-aliasing traditional rasterizers using deep learning as well, which will probably be practical sooner than the raytracing/denoising stuff. This brings me great joy and satisfaction.
|
# ? Jun 19, 2018 10:16 |