Research Scientist at Google DeepMind (previously Ph.D. @ Harvard) doing computer vision. I also like music.
Joined October 2016
- Tweets66
- Following308
- Followers773
- Likes317
Dor Verbin retweeted
Introducing beignet, a new neural field architecture!
beignet uses a multiresolution fourier frequency pyramid to achieve faster convergence and higher precision when solving PDEs and finding singularities.
Teaser: here is beignet at work fitting an image (super fast)!
Dor Verbin retweeted
Introducing Generative Commerce.
@daletlabs – at New York, NY
We released the code (and weights!) for our model, so you can finally try it out. Also if you're attending SIGGRAPH this week you should check out Xiaoyan's talk on Wednesday!
See project page for more details: gr3en-relight.github.io
Finally, GR3EN is open source! 🎉
Click the lights in any video, choose their colors, and GR3EN relights the entire scene with consistent shadows and reflections. It works on real-world videos—including robotics and egocentric videos—directly in your browser, with zero setup. 👇
🤗 Demo: huggingface.co/spaces/xyxing…
📄 Paper: arxiv.org/abs/2601.16272
💻 Code: github.com/xyxingx/gr3en
🌐 Project page: gr3en-relight.github.io
🎤 We’re presenting at #SIGGRAPH2026 this Wednesday, July 22:
Oral: 2:10 PM
Poster: 3:30 PM
Room 408A
Come say hi!
Dor Verbin retweeted
📢📢📢introducing 𝐏𝐨𝐰𝐞𝐫 𝐅𝐨𝐚𝐦
A 3D representation that can be ray traced or rasterized in real time, with NO COMPROMISE in quality.
- Project: powerfoam.github.io
- arXiv: arxiv.org/abs/2604.24994
Rasterized at 3DGS-class FPS
Ray traced at Radiant Foam speeds
Check out our work on 3D scene relighting! We rendered purely synthetic data and used it to fine-tune a video model for relighting. Even though it was only trained on synthetic data, it generalizes to real captured scenes—shadows, reflections, and all!
[1/6] We introduce GR3EN, a generative approach for relighting 3D environments. It gives you full control over every light source in a 3D scene — enabling adding new light sources, changing the color of existing ones or turning them off entirely. Shadows, reflections, and interreflections all update realistically.
Accepted to SIGGRAPH 2026 🔥
Project_page: gr3en-relight.github.io/
From: @GoogleDeepMind @GoogleResearch @UvA_Amsterdam
We'll be presenting our paper at SIGGRAPH this year!
See more results and details about our work at:
- Project page: gr3en-relight.github.io/
- Paper: arxiv.org/abs/2601.16272
Dor Verbin retweeted
Our short film Dear Upstairs Neighbors is previewing at @sundancefest. 🎬
It’s a story about noisy neighbors, but behind the scenes, it’s about solving a huge challenge in generative AI: control.
Developed by Pixar alumni, an Academy Award winner, researchers, and engineers, here’s how it came together. 🎨
Dor Verbin retweeted
📢📢📢 𝐒𝐩𝐡𝐞𝐫𝐢𝐜𝐚𝐥 𝐕𝐨𝐫𝐨𝐧𝐨𝐢: directional appearance as a differentiable partition of the sphere
– Project: sphericalvoronoi.github.io
– Paper: arxiv.org/abs/2512.14180
– Colab: tinyurl.com/spherical-vorono…
Project led by 𝐅𝐫𝐚𝐧𝐜𝐞𝐬𝐜𝐨 𝐝𝐢 𝐒𝐚𝐫𝐢𝐨 (graduating!)
We're running a workshop on physics-inspired 3D vision and imaging at CVPR this year! We'll have a poster session, so if you'd like to present something, please submit your work by the April deadline. Otherwise, just come hang out with us at the workshop! pi3dvi.github.io
📢📢📢 Come and submit to our workshop on Physics-inspired 3D Vision and Imaging at CVPR 2025!
Speakers 🗣️ include Ioannis Gkioulekas, Laura Waller, Berthy Feng, @SeungHwanBaek8 and @GordonWetzstein!
Thanks to coorganizers @imarhombus, @ceciliazhang77, @dorverbin and @jtompkin!
Dor Verbin retweeted
The code for EVER: Exact Volumetric Ellipsoid Rendering has finally been released!
To recap our paper, EVER is an exact, real-time volume rendering method, capable of achieving the highest quality results on large scenes!
Code: github.com/half-potato/ever_…
Dor Verbin retweeted
🚀 Introducing SimVS: our new method that simplifies 3D capture!
🎯 3D reconstruction assumes consistency—no dynamics or lighting changes—but reality constantly breaks this assumption.
✨ SimVS takes a set of inconsistent images and makes them consistent with a chosen frame.
We’ll be presenting NeRF-Casting at SIGGRAPH Asia next week! NeRF-Casting enables photorealistic rendering of scenes with highly reflective surfaces—something that was previously impossible with models like Zip-NeRF and 3DGS. (1/6)
A particularly cool feature of our method is that it can recover the geometry of reflected objects that aren’t directly visible in any of the training images—only seen through reflections, like this lamppost reflected off the hood of the car and its right windows! (5/6)
The full paper and a lot of results are on our project page: dorverbin.github.io/nerf-cas… (6/6)
Our team is looking for interns! If you're interested in working with us next summer, please apply.
Dor Verbin retweeted
(1/N) Flash Cache: Reducing Bias in Radiance Cache Based Inverse Rendering
Website: benattal.github.io/flash-cac…
tl;dr our #ECCV2024 (oral ✨) paper presents a new system for inverse rendering that is more physically accurate, and therefore less biased, than existing approaches.
Dor Verbin retweeted
Our new paper performs exact volume rendering at 30FPS@720p, giving us the highest detail 3D-consistent NeRF!
Paper: arxiv.org/abs/2410.01804
Website: half-potato.gitlab.io/posts/…