Tomas

Fake Satellite Imagery This episode is all about using AI to generate satellite imagery using something called generative adversarial networks ( GANs ). mapscaping.com/podcast/fake-s… Can you spot the fake image?

Riffusion, real-time music generation with stable diffusion @huggingface model: huggingface.co/riffusion/riff… project page: riffusion.com/about

We are excited to announce the release of Stable Diffusion Version 2! Stable Diffusion V1 changed the nature of open source AI & spawned hundreds of other innovations all over the world. We hope V2 also provides many new possibilities! Link → stability.ai/blog/stable-di…

We are launching a new arXivLabs collaboration with @HuggingFace to make demos related to papers in cs, stats, and eess directly accessible from arXiv!

90 Days of Diffusion 90 🤯 AI Advances 👇

I've trained a latent diffusion upscaler for the Stable Diffusion autoencoder (and anything you feel like feeding into it if you can tolerate a little artifacts) in collaboration with @stabilityai. Try the Colab written by @nshepperd1 here: colab.research.google.com/drive/1o1qYJcF…

Glad to share our #WACV2023 paper with @jormalaaksonen @Xelfor @heikkiarponen @NotACyborgYet @CSAalto! Problem: Unaligned vision(V)-language(L) pre-training. Idea: Generate some imaginary V from L, then L can interact with its imaginary friends 😃 arxiv.org/abs/2210.13591

1 week of Stable Diffusion A creative explosion is unfolding with Stable Diffusion,s showing the power of open source as state of the art! We curated 23+ applications this week: new features, workflow integrations, UIs; run on Win, CPU, AMD, M1 and more! multimodal.art/news/1-week-of…

We took a representative set of 5.6 billion images from the internet, filtered out weird and low quality stuff to 2 billion, 100 terabytes, and squished it to a 2 Gb file we are making available for anyone to use versus monopolising it.

When people ask what I do #dalle2

A colleague of mine generated these Dall-E 2 remote sensing synthetic images, they are pretty convincing.

Most machine learning libraries haven't been designed to work with geospatial data. #TorchGeo, a PyTorch domain library, is set to change this by monitoring some of the world’s greatest challenges, like natural disasters and climate change. Read more: bit.ly/3tT6Mw7

CogView2: Faster and Better Text-to-Image Generation via Hierarchical Transformers abs: arxiv.org/abs/2204.14217 CogView2, shows very competitive generation compared to concurrent state-of-the-art DALLE-2, and naturally supports interactive text-guided editing on images.

ARF: Artistic Radiance Fields abs: arxiv.org/abs/2206.06360 project page: cs.cornell.edu/projects/arf/ github: github.com/Kai-46/ARF-svo… create high-quality artistic 3D content by transferring the style of an exemplar image, such as a painting or sketch, to NeRF and its variants

🎉 Happy to share a blog @krasul and I have been working on: the Annotated Diffusion Model. We implement and train Jonathan Ho et al's DDPM (that forms the base of today's DALL-E 2 and ImageGen) step by step in PyTorch: huggingface.co/blog/annotated…

“Finish the cat drawing” viral meme tweet has replies with all sorts of nice, creative ‘out of the box’ thinking. I use #Dalle’s inpainting function to do this task, and was impressed at what it can do. Here is the output using the prompt “cats” 🧵An entire thread of results 🐈

Following Imagen example, I plotted the Pareto CLIP score vs FID over conditioning guidance scales on dalle-mega. We can notice a few interesting things.

Let's play a game. Two cat photos, one is real and the other is a fake generated by DALL-E. Can you figure out which which? Round 1: