Excited to share our #ICCV2025 work Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets!
Our method generates large sets of images using significantly less compute than standard diffusion.
📎https://ddecatur.github.io/hierarchical-diffusion/
1/
After a prolific and rewarding decade at Adobe Research (see vova.kim), I’ve taken on a new role as Director of Applied AI/ML at Adobe. I’m building a new team and we’re looking for outstanding Applied Scientists (mostly for visual data). Shoot me an e-mail or pass along!
Nice! Do you want to do an AGI startup together?
Congrats!
In other words:
1. “Nom-nom yummy minerals” =>Shakespeare/da Vinci [impressive emergent behavior]
2. Millions of books and images => coherent sentences / images [interesting, but far less impressive]
Artificial intelligence is hand-crafted and fine-tuned to process human abstractions. So its ability to handle them well does not seem as surprising.
A remarkable thing about human intelligence, is that a single-celled slime, driven purely by natural selection, somehow turned into a hairless ape with a seemingly useless adaptation which allowed it to build generalizable abstract representations of the natural world. AI, on the other hand..
Code available on Valentine’s Day? :)
Yes!
Why would these entities value their survival? There is no objective function or evolutionary pressure for them to develop this kind of behavior.
🎓 We introduce SAMa! A material selection and segmentation model on 3D models in any format (3DGS, NeRF, Mesh).
Given a user click, we propose to select all regions on an objects with the same material. We can also do segmentation in under a minute: mfischer-ucl.github.io/sama/
Took us 3 years to get a decision back, one reviewer complained that we didn’t cite 3 papers from this year (they were published 2 years after we submitted)
🌟 Text-to-3D is awesome ! But how do we iterate on the generated 3D model, to get just the right result? Do we tweak the prompt endlessly? Revert to traditional 3D modeling techniques?
We propose a solution to "3D inpainting” 🤩🎨
Project: amirbarda.github.io/Instant3dit....
A thread. 🧵 1/
