Gene Kogan
@genekogan
Followers
26,537
Following
778
Media
905
Statuses
5,098
My avatars @eden_art_ @abraham_ai_ @mars_college
Joined June 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
#母の日
• 224389 Tweets
#MothersDay
• 137704 Tweets
ヴィクトリアマイル
• 131981 Tweets
ストフェス
• 58861 Tweets
テンハッピーローズ
• 53769 Tweets
Chivas
• 43628 Tweets
カーネーション
• 39961 Tweets
Roger Corman
• 37014 Tweets
ナミュール
• 35804 Tweets
マスクトディーヴァ
• 30546 Tweets
Loma
• 25718 Tweets
WIN5
• 25648 Tweets
フィアスプライド
• 19360 Tweets
津村騎手
• 16939 Tweets
ツムツム
• 15495 Tweets
#SHOWチャンネル
• 13788 Tweets
ヴェルディ
• 11303 Tweets
G1初制覇
• 10779 Tweets
ライラック
• 10693 Tweets
ウンブライル
• 10261 Tweets
just in case you were wondering, this is what happens when you repeatedly run FaceApp’s smile filter until it can't detect a face anymore
111
1K
3K
Text-to-image with OpenAI's CLIP is getting more and more surreal. Here is "A banquet hall"
28
380
2K
NVIDIA's new vid2vid is the first open-source code that lets you fake anybody's face convincingly from one source video. prior "face2face" stuff was either cartoonish or proprietary. interesting times ahead...
35
965
2K
very cool demo of adjusting attributes of natural scenes interactively (season, time of day, sunlight/cloudiness). code said to be coming soon. this is the photoshop of the future
13
460
1K
The new StyleGAN code is super neat. You can cross-breed latent vectors to get a sort of style transfer effect.
9
276
1K
texture synthesis infinite loop, based on The Great Wave off Kanagawa (Hokusai, ~1830)
13
270
968
The best thing about making generative art for a living is that sometimes buggy code works better than good code.
23
86
956
progressively-grown GAN (Karras et al) trained on ~80,000 paintings
12
217
945
imitating videos with BigGAN: first predict Y from the source, then feed it to the generator. input here is the trailer for Planet Earth
16
214
902
the future of AI-generated video game stock characters looks fine
23
152
807
groovy! i figured out how to loop dreams. watch carefully... this video is only 3 seconds long ♻
9
173
800
Looping text-to-image of "A drawing in the style of M.C. Escher"
8
129
726
upgrading meat puppet to HD. as soon as i can get a good voice model, i’m going to make a parallel state of the union address
14
137
696
C++ port of Whisper, very impressive. Base model does faster-than-real-time speech-to-text on CPU on a Macbook Pro
15
109
699
Successfully replicated NVIDIA's 512x512 labels-to-landscapes model (GauGAN) for use with SPADE. The model can be downloaded from
It will also be available on
@runwayml
shortly. Here it is in action!
8
134
577
you can now visualize layers & classify images ~60fps w/ ofxDarknet. real-time feature extraction in
@openframeworks
4
205
532
walking the latent space along b-splines instead of linear interpolations (zig-zagging). the effect is subtle, but it rounds the corners a bit smoother
10
73
499
CycleGAN "failure cases" are as interesting as the successes. ML papers need more blooper sections
11
237
490
8 years ago today, I posted this style transfer experiment of a scene in Alice in Wonderland ()
It was my first "generative AI" (although no one called it that at the time) work to go viral, getting 100k views in a few days. It convinced me to focus on…
18
54
488
The Neural Aesthetic is finished! Notes and around 30 hours of video lectures are up here:
A big map of all the slides can be browsed here:
7
119
470
Sometimes if I want advice about something, I start drafting a long message to someone about it. Writing it down forces me to think so clearly that I end up answering my own question, and never end up sending it.
The lesson is: write! Writing is an upgraded version of thinking.
12
37
457
This cartoonization code makes basically anything look like anime
19
80
460
if you think AI hype is bad today, read this take from the NY Times in 1958 after perceptrons were invented
10
137
441
neat! an open-source clone of SimCity 2000. i loved this game as a kid. once the simulation is implemented, will be a great playground for RL agents
9
166
433
nice of NVIDIA to open-source the models for their progressive GANs. it's a treasure trove
5
115
431
really excited to have DarkNet in
@openframeworks
. here's dense-captioning at 2fps on my laptop. lots of potential
ofxDarknet: realtime gpu accelerated neural networks in oF. /w help by
@genekogan
and
@bildpeter
@ofxaddons
6
74
239
5
148
433
working on an "aging mirror" that fast-forwards you ~10-20 years into the future. works in reverse as well! (nice cameo by
@sam_hains
)
10
60
420
StyleGAN was already pretty good at faces. The improvements in v2 are really visible in the other classes
#StyleGAN2
2
69
407
not every day you get to see your work up on a never-ending LED screen. thanks
@digilogue_
for inviting me to Istanbul
7
36
407
If you're disillusioned with Siraj and looking for video channels with high-quality, accessible & authentic AI coverage, I would go to
@karoly_zsolnai
's "Two Minute Papers" and
@xsteenbrugge
's "Arxiv Insights"
12
49
383
student projects at my workshops get more and more sophisticated every year. this week at
@ecal_ch
, Nathan Vogel wrote a firefox extension which intercepts OpenStreetMap tiles in realtime, sends them to CycleGAN to be turned old-time style, then re-injects them back into the page
7
51
383
How neural networks are trained:
@ml4a_
's longest & most advanced chapter, from linear regression to backprop, is finally drafted. feedback & proofreading is welcome
5
123
370
1/ New essay: an excavation of the history of mathematics, from deep learning down
5
151
364
some more neural synthesis/deepdream experiments... masks for new channels radiating outward from the center
8
77
367
demo notebook in colab for stringing together interpolations from the newly released
#BigGAN
models. have fun :)
3
92
365
hi-res texture synthesis, i.e. contentless style transfer. here is one based on Kandinsky's Composition 8 (1600x1260 version at )
6
51
355
Training a sketch-to-satellite model with SPADE. It generates satellite imagery of Berlin from hand-drawn maps
10
78
347
adversarial generation of 3d point clouds from Achlioptas et al. very nice latent space. seems like 3d is on its way
4
132
341
I'm very excited to share my first NFT on
@withFND
, now freshly minted.🌿
This is one of my personal favorite works. ❤️
#NFT
#nftart
#Cryptoart
30
36
337
I mixed up the golden rule and golden ratio and started treating others 1.62x how I wish others to treat me.
2
50
338
CariGANs -- drawing caricatured faces, like those street artists around Times Square. goes in reverse too. very cute idea (╹◡╹)凸
4
81
326
run fast-style-transfer on your webcam. including models for kandinsky, hokusai, picasso, lichtenstein & google maps
5
80
321
Trying to animate latent visions. Going for something a bit smoother, but I’m enjoying the stop-motion like feel of my early attempts at it.
“Alone on the water”
12
25
327
Turing’s “Can machines think?” is much more often quoted than his “the question is too meaningless to deserve discussion” a few pages later
3
161
306
Here's what a few years of progress on text-to-image generation looks like, one prompt at a time.
"Frank Sinatra as a purple alien in surrealist style"
1) AttnGAN (2018)
2) CLIP+VQGAN (2020)
3) CLIP+Diffusion (2021)
4) DALL-E 2 (2022)
7
85
313
it's like a Shepard tone. the center mask appears as if it's cascading outwards, but it's an illusion. it always stays in the same place
8
96
309
SampleRNN generated death metal on 24/7 livestream. that there's no humans and it's infinite is the most metal thing ever 🎸💀🔥🤘
7
115
299
New York!! I'm looking for an apartment starting Oct/Nov. Know any place that looks like this? Please tell me about it :)
11
41
293
navigation in a nightmare: the map appears to zoom-in forever but never get anywhere
6
65
294
3D photographic inpainting from a single source image
Adds a whole new dimension to these puppyslugs
12
56
281
very cool work on text attribute transfer, e.g. sentiment (positive->negative), genre (humorous->romantic), etc... like pix2pix for text! + code at
3
87
284
massive new dataset of 1M+ indoor scenes with texture, lighting, depth/normals, semantic maps, and more. biggest of its kind that i know of. with a bit of pre-processing, this could be used to train 3d generative models (of just furniture, but still neat)
2
81
272
CLIP-guided image transformations are really cool. A hint of what Photoshop might look like in the future...
Input: a picture of mountains
Output: input + "An ukiyo-e painting"
3
34
267
here's a sneaky way to get applause. make the crowd play the T-Rex game with claps 😂
2
50
262
4.5 years of GAN progress, visualized to scale (modified from )
3
68
263
The
@runwayml
beta is public! We're very excited to finally have it out in the open. Give it a try and let us know what you think.
7
62
258
some fun new demos combining
@p5xjs
with tensorflow.js.
make music with PoseNet:
play pong with your camera:
7
55
248
Trying to reproduce NVIDIA's unreleased labels-to-landscapes model (GauGAN). One cool result after 20 epochs is it renders reflections of stuff over surfaces of water, even though there's nothing in the labels
5
44
250
very excited to announce i'll be back in New York this fall, starting a residency at
@ITP_NYU
and rebooting my machine learning for artists course. draft syllabus at
10
31
246
I've made a PyTorch-based fork of neural-style which adds spatial control when using multiple styles (aka masked style transfer). Code is here:
6
37
248
Trying out the new StyleGAN's encoder, projecting real faces into the ffhq model.
3
27
241
I always thought of Schmidhuber as an artist. His emphasis on authenticity, and willingness to write about unusual topics with creativity and rigor at the same time, e.g.
Awards don't deserve him.
2
37
240
My first NFT now available on
@rariblecom
! 🥳
It is a looping texture synthesis based on Hokusai's iconic Great Wave off Kanagawa. It would be the first artwork I've ever sold. I hope it finds a nice home.
#cryptoart
#NFT
13
24
237
There is a fake news generator on Colab. Seems as good or better than GPT-2 to me. The prompt is the top headline from the NYTimes, first attempt. Try it out
9
64
225
over the past two years, i've been very slowly writing a short creative writing piece about the history of mathematics, and i think it's finally done. just proofreading it and posting tomorrow. a hint...
9
16
225
Open call!
If you're into:
- nature ☀️🌵🏜️
- art 🎭🎨🎵
- code 💾🤖💻
- yoga 🧘🧘🧘
- building ⚒️🏚️⚙️
- gardening 🌺🌱🌻
I'm helping organize a free & highly unconventional experiment in living & learning in the desert, 2020. Get in touch if interested!
8
44
225
A short introduction to the
@abraham_ai_
project to study and build an autonomous artificial artist
6
63
222
Introduction to generative models:
Eigenfaces, autoencoders, GANs, ++ tutorials on dataset collection and training your own DCGAN.
Week 6 video and slides
2
55
222
The goal of the
@abraham_ai_
project is to create an autonomous artificial artist, an agent that makes unique and original artworks through the decentralized curation and governance of AI-based generative art.
20
30
213
updated my fork of the
#BigGAN
colab notebook with a GUI for easily making A+B mashups and interpolations, auto-saving the images locally or to gdrive. post your best mashups!
2
48
220
The tentative idea is to make this "multi-player" and integrate it into
@abraham_ai_
as a co-creation/co-curation tool.
6
10
220
running face recognition over that one-hour video of GAN-generated celebrity faces released by
@NvidiaAI
last week ()
6
61
217
neat work (+code) generating video snippets with GANs. the babies are terrifying though
2
118
207
Videos and slides for my first two classes at
@ITP_NYU
this semester are now up.
1) The whole class "in 60 minutes":
2) Neural networks:
2
54
208