Working on a Unity3D (inactive) <> Runway GAN rendering plugin. Generate semantic maps from 3D scenes in real-time, use them as input to semantic image synthesis models like SPADE, pix2pixHD, etc.

This is a great update! Can't wait to try it out!!! 👏👏👏

A wonderful article about training your own #StyleGAN model with lots of tips and tricks included! gwern.net/Faces #creativeAI

I don't always experiment with an #AI, but when I do I am wearing this T-shirt! 😃

Today, we had a nice session with ALTTAB: platform for critical and #speculative design in Prague, introducing Runway, sharing some experiments and conceptualizing the visual outputs for a speculative use-cases. More workshops to come! w/ Pavol Rusnak

We propose InterFaceGAN to interpret the semantics inside the pretrained face synthesis GANs. Various facial attributes emerge spontaneously as disentangled subspaces in latent space, even for pose and synthesis quality. Paper and code are available at shenyujun.github.io/InterFaceGAN/

I've just published a new work on my website, using an AI to talk in vector : fluate.net/en/travaux/vec… Vectoglyph - Vector forms as a foreign language for the artificial intelligence GPT-2 #language #vectorart #aiart #gpt2 #openai #artificialintelligence #ai #neuralnetworks

Oh My #GauGAN this worked better than I expected. 🤔 #SIGGRAPH2019 @NVIDIADesign

I created a #GIMP plugin so you can play with #GauGAN in a fully featured graphics editor. @NvidiaDesign GIMP #SIGGRAPH2019 github.com/prusnak/gimp-p…

For a month I've been working on an open source tool for labelling images in a browser - here it is makesense.ai Vincent Mühler Ben Hamner PyImageSearch Lavanya 🐝 🦋 shiffman.lol Maybe you'll find it useful in the next ML/CV project? visit GH: github.com/SkalskiP/make-…

Training a sketch-to-satellite model with SPADE. It generates satellite imagery of Berlin from hand-drawn maps

I launched Selfie2Anime with Rico A. Beti, a website where you can find out what you look like as an anime character! 😄 Have a look! selfie2anime.com #selfie2anime

Dear friends, today I spent around 6 hours in our latest GAN training for the City of New York, Machine Hallucination. I found epic latent spaces that I would like to share with you! ARTECHOUSE

ImageJ is kind of like Runway ML for cell biologists. Segmenting cells, content-aware fill for fluorescence microscopy, etc. You can use it for whatever you like, of course...

What we've been all waiting for is here! Photoshop meets Runway ML 🥳

Another big improvement on semantic image synthesis, code promised soon. Results look amazing compared to SPADE, but took 16 Titan X (!) to train arxiv.org/abs/1910.06809

Remember thispersondoesnotexist.com? Together with Lenka Hamosova we launched strangeattractions.xyz showing what #StyleGAN dreams about. Here's the blog post that goes a little bit more into the detail how this was born: rusnak.io/strange-attrac… #creativeAI #GAN

Colab integration with Google Drive just got better. Authenticate only once per notebook. No authentication codes are necessary for private notebooks when using the 'Mount Drive' button in the file browser.

What if I told you that your face will soon be deepfaked daily, and these deepfakes will be shown to your family, colleagues, and friends? Let me introduce the speculative scenario of personalized synthetic advertising 🤯 👉 link.medium.com/ZRGoBHkm07