Our social media team hosted by @antoninofurnari.bsky.social @natanielruiz.bsky.social @yaelvinker.bsky.social @csprofkgd.bsky.social. Excited to bring you the news and announcements this #ICCV2025! Special thank you to the Organizing Committee and their great work as we head into the main program.

Tomorrow's day begins at 8:00am sharp with some amazing oral sessions, followed by Prof. Smith's keynote at 9.30am in Exhibit Hall III. Join us for both! ✨✨✨ The efficiency of learner generated experiences - Linda B Smith abstract: iccv.thecvf.com/virtual/2025...

Reviews released to authors soon. For more information please check the official website iccv.thecvf.com/Conferences/...

imagen 3 pretty useful for adding cuteness to the deadline posts

if you're submitting to #ICCV2025 please double check this list and remember to not include identifying information :)

Starting with our daily reminders of the final paper submission and supplementary materials deadline for #ICCV2025 which is on March 7th, 2025. Please ensure to upload large supplementary material well in advance of the deadline to avoid issues.

🥳

Collected all the original D&D books and the original supplements 🤩

fal.ai is my favorite way to generate images

I do actually feel sorry for the guy who accidentally binned his b*tcoin hard drive, just because the psychic damage you would take from that is immense, and even the most sane of us would likely ten years later still be trying to scheme a way to dig it up

beautiful post by Ruiqi and other amazing researchers

really cool work on story-to-video generation link: dreamrunner-story2video.github.io

Introducing Generative Omnimatte: A method for decomposing a video into complete layers, including objects and their associated effects (e.g., shadows, reflections). It enables a wide range of cool applications, such as video stylization, compositions, moment retiming, and object removal.

interesting paper on test-time training for solving ARC. a cool fact is that dreambooth can be considered as test-time training for subject driven generation. test-time training is very powerful t.co/OzbUEthhSo

Around two years ago we made the first comic fully generated by an AI with a consistent character using DreamBooth. The character is Anselmo, which comes from drawings from a friend from my hometown in La Paz, Bolivia

1/1

labs already replicating o1 success

big news for Flux by Robin Rombach

great starting pack by Ariel

google did the "open ai one-up" to the open ai one-up

other methods vs. our magic insert

cleave {to} me

check out this starter pack by Sander!

Another AMAZING addition to my academic library. @phillipisola.bsky.social is here too!

Sharing a work I was recently supporting called M-VAR. The future of generative vision might be autoregressive! M-VAR is an evolution of the VAR model that is significantly more efficient to train and run. 🧵 web link: oliverrensu.github.io/project/MVAR... code: github.com/OliverRensu/...

crazy to think dreambooth has 2k+ citations

bot problems have gotten to an all-time high on twitter and the company does nothing to remedy. it seems like things are better around here for now

my goal is to be here several times a week and post our AI/ML and generative model work at Google that we've developed in the last years, as well as post some other work that I like. kinda like a greatest hits ☘️

hello world (follow me if you like generative models)