Scott Eaton

Experiment with Geometry and Motion, variation II

@mkturkcan great work. potentially very useful for a multi-class detection requirement we have. Excited to test it. 👍

Corrected link to code: github.com/mkturkcan/DART

Excited to share DART: Detect Anything in Real Time! We rebuilt SAM3's SOTA detection pipeline for multi-class real-time deployment. 15.8 FPS with the full ViT-H, 37.6 FPS with a pruned backbone, 55+ FPS with distilled students. Any vocabulary, on a single GPU.

@cogspa @EpicArtFail @linuz90 @LinusEkenstam Thanks Joe. Quickly - my work is trained exclusively on my own data: photography and CG. It pre-dates 'internet scale' models (Stable Diffusion, MidJourney, etc). ML is a powerful tool when understood and used imaginatively, but I will always side with artists WRT copyright.

You may call it texture synthesis, and you aren't wrong, but he calls his work "creative AI" and he states he is using Machine Learning. He spent a 3 year residency at META. scott-eaton.com/category/creat… He is using his data, which is ethical as far as I know. I've been a fan of work for many years. No, he is not using Midjourney or Stable Diffusion. I believe this is 100% him, from his drawing to his coding, but I would love to hear him comment on it. He is an amazing artist and educator. If you are reading this, Scott, I appreciate your work and share it with my students at PCC. scott-eaton.com/experiments

AI haters will always hate. Here is daft punk “stealing” 25 years ago.

Layers of movement captured in abstraction: @_ScottEaton_ ’s Contemplating Motion is like watching AI learn to dance with memory and light. An ethereal play of forms, fading in and out like echoes of thought 👀 @FeralFile

the secret is in the negative space.

@ChngsungJ89002 Yes, I believe so. 👍

@_ScottEaton_ quick question, is this book shown in these links, an legit exact official translation of Die Gestalt des Menschen? archive.org/details/got-bam amazon.com/Complete-Guide…

@zachlieberman love this one Zach.

@deviparikh Wow, big news. Best of luck with the next thing Devi, I am sure it will be great!

Update: I left Meta yesterday. After 7.5 years. I am sad, nervous, and excited. Sad because I'll miss Meta! I've felt tremendously valued my entire time at Meta (first in FAIR and recently in GenAI). I'll miss the people and being in the thick of things. Nervous because who in their right mind walks away from the job I had in times like these (leading research efforts in generative media and multimodal LLMs)?! And excited for new experiences :) Stay tuned for when I have more to share!

Also running a drawing workshop at the VIEW Conference @ViewConference (tomorow). Two hours of drawing in the sketchbook, good old-fashioned fun.

Excited to be giving a talk at the amazing @ViewConference titled "Out of Distribution - Thriving Creatively in the Age of AI" (viewconference.it). The talk covers a lot of ground, but essentially it's about how we, as artists and creatives, need to think about situating our work and careers in relationship to this new, relentlessly disruptive technology. I also share many examples of my AI/ML experiments, including a behind-the-scenes look at my recent takeover of Times Square. Hope to see you there! Special thanks to @Lenovo and @nvidia for sponsoring this talk.

Learn & be inspired by a legend at @ViewConference. Don’t miss Dynamic Figure Drawing with @_ScottEaton_ Join us! viewconference.it/pages/registra… #viewconference2023 15-20 Oct, Turin, Italy #ScottEaton #art #drawing #ai #cgi #vfx #animation #games #realtime #vr #xr #mr #3dart #film

@theAliceRoberts @TimesSquareNYC Thanks Alice!

@_ScottEaton_ @TimesSquareNYC I love it! Reminds me of the architecture of trabecular bone too...

My new experimental piece, Intersections, takes over the screens in Times Square, NYC @TimesSquareNYC for the entire month of August (3 minutes to midnight, every night). If your in the city and up late, check it out. More here: scott-eaton.com/2023/intersect…

Excited to be giving a talk for @Google today titled, "Inside Pandora's Box: Bodies, Geometry & Internet-scale Intrigue". Lots of images, videos and reflections on my last six or seven years exploring machine learning / AI in my work.

@ViewConference announces that @_ScottEaton_ is a confirmed speaker attending in person. Meet your heroes! #viewconference2023 “The Future We Want” 15-20 Oct, Turin, Italy Sign up: viewconference.it/pages/registra… #viewconference #ScottEaton #aiandart #generativeAI #art #ai #film

@IrinAngles @NeurIPSConf lovely pose!

@_ScottEaton_ @NeurIPSConf I love this statue so much because it reminds me of myself. How can we have the same ideas but at different times Life is so interesting 🧐

My @NeurIPSConf talk for the "Machine Learning for Creativity and Design" workshop is online here: slideslive.com/38938166/artis… I talk about how I incoporate machine learning into my artistic practice, show the development of recent work, and also cover some best practices.

Panel Discussion on creative AI: Luba Elliott @elluba , Anna Ridler @annaridler & Scott Eaton @_ScottEaton_ 🎙️🙌🏻

We are delighted to host a panel discussion on creative AI with artists @annaridler and @_ScottEaton_, moderated by AI expert, @elluba. Please join us on Monday 20 March, 6 - 7:30pm GMT RSVP 👇 eventbrite.com/e/panel-discus…

A brand new @beforesmag podcast is here! It's @_ScottEaton_ and his in-depth take on generative art, Midjourney, A.I., neural networks, machine learning and what it means for artists. beforesandafters.com/2023/01/09/sco…

a few more from the same. Where the mask and prompt are not aligned, the semantic mask still gives very useful direction to the generator, but doesn't overly constain it. Here exploring abstract possibilities. #MakeAScene @MetaAI

Interestingly, if the text prompt is entirely unrelated to the categories prescribed in the label, Make-A-Scene uses the label only as loose guidance for compositional layout. Here are a few examples where the text prompts were wildly different from 'bowl,mirror,vase' categories.

Excited to share a few additional experiments from @MetaAI 's Make-A-Scene. The compositional control is a powerful augmentation on top of the imaginative capabilities of these text-to-image models. These are all from the same semantic layout, just prompt variations.