Discover and read the best of Twitter Threads about #CLIP
Most recents (11)
π #StableDiffusion can be fine-tuned to generate medical images, and the outputs can be controlled using natural language text prompts!
In our latest work, we use SD to create synthetic chest xrays and insert pathologies like pleural effusions.
𧡠#Radiology #AI #StanfordAIMI
In our latest work, we use SD to create synthetic chest xrays and insert pathologies like pleural effusions.
𧡠#Radiology #AI #StanfordAIMI
#stablediffusion is a #LatentDiffusionModel and performs its generative tasks efficiently on low-dimensional representations of high-dimensional training inputs. SD's VAE latent space preserves relevant information contained in CXR; they can be reconstructed with high fidelity.
#StableDiffusionβs output can be controlled at inference time by using text prompts, but it is unclear how much medical imaging concepts SD incorporates. Simple text prompts show how hard it can be to get realistic-looking medical images out-of-the-box without specific training.
I discovered a bug in my own Diffusion + CLIP pipeline and suddenly the samples are unreal.. π€―
Here's
"Just a liquid reality..."
#AIart #notdalle2 #Diffusion #clip
Here's
"Just a liquid reality..."
#AIart #notdalle2 #Diffusion #clip
"The magnificent portal of mother Gaia"
"Framing reality"
The idea of building an autonomous entity that is able to be creative or at least co-creative in some form goes back hundreds of years. Literature is full of examples of machines or contraptions that could be seen as some precursor of the idea of AI:
frontiersin.org/articles/10.33β¦
frontiersin.org/articles/10.33β¦
I guess anyone who starts with generative art quickly comes to the question: "how can I automate this fully?". For me that moment came when I wrote my first "try out every possible pixel combination" program back in 1985 or so on my Commodore C-64.
Of course I quickly had to realize that brute-force does not get you anywhere and you need more intelligent approaches to separate the interesting from the noise or the banal. So I had to wait for more than 20 years for AI to get to the point where it became feasible.
Note to self: don't use default matplotlib colormaps to make digital artπ€¦ββοΈπ
New samples from my 'color-quantized VQGAN' are looking great!
Here's "π¨πππππ πππ ππ πΎπππππππππππ, π πππππππ ππ π πππ ππ ππ πππππππ"
#clip #AIart
New samples from my 'color-quantized VQGAN' are looking great!
Here's "π¨πππππ πππ ππ πΎπππππππππππ, π πππππππ ππ π πππ ππ ππ πππππππ"
#clip #AIart
"ππ ππππ
ππ ππππ ππ πππππ"
"π΅ππππππππ"
I slightly patched the @dribnet's amazing CLIPIT+PixelDraw project and added two new features: a palette enforcement and a smoothness enforcement.
Check some examples in this thread. First, [deep ocean monsters] with the blue palette enforced. #clip #pixelart
Check some examples in this thread. First, [deep ocean monsters] with the blue palette enforced. #clip #pixelart
"Imagination is found to be leading us inexorably toward our objective - toward a fulfillment of our desires." #clip
"You got to sit down with people and talk about the material they put into the film." #clip
"You almost can't believe in god any more than believing in good people or in dolphins." #clip
1/ What makes this @ChemicalScience front cover so special is how it was made.
It is probably the first-ever journal cover designed using a VQGAN + CLIP and the unreal engine trick!
#VQGAN #CLIP #GenerativeArt
More information in the thread β¬οΈ
It is probably the first-ever journal cover designed using a VQGAN + CLIP and the unreal engine trick!
#VQGAN #CLIP #GenerativeArt
More information in the thread β¬οΈ
2/ The AI model (VQGAN + CLIP) generated most of the image using βenzymatic chemical reactions. green chemistry. advanced unreal engineβ as input.
Itβs interesting that you can recognise the lab with the blackboards, the floor and the βreactionsβ.
Itβs interesting that you can recognise the lab with the blackboards, the floor and the βreactionsβ.
3/ Iβve added an enzymatic reaction from our publication (pubs.rsc.org/en/content/artβ¦) on top of the image.
Before (AI generated) | after (final cover)
Before (AI generated) | after (final cover)
El @INEMexico decidiΓ³ cancelar el registro de @MexLibre_ por el #clipgate y $1,241,687 equivalentes al 8.18% de su financiamiento cuyo origen βno pudo identificarβ con la mofa gubernamental e ilegal correspondiente del tema Abro 𧡠1/22
En su defensa @CiroMurayamaINE seΓ±ala que le hizo una prevenciΓ³n a @MexLibre_ desde Agosto/2019 por $7,604.10 que @FaustoBarajas ya dijo cΓ³mo la resolviΓ³, y de ahΓ surge la cifra de $1,241,687 del 4/septiembre; comparto ahora datos del dictamen del @INEMexico INE/CG638/2018 2/22
En Γ©l se seΓ±ala que cuando @lopezobrador_ fue Dirigente de @PartidoMorenaMx, el fideicomiso βPor los demΓ‘sβ constituido 25/Sept/2017 para apoyar a damnificados de los sismos tuvo ingresos por $78,818,566.37, y de esos $44,407,082.49 fueron en EFECTIVO, ni a Clip llegaron! 3/22
