With Karan Ahuja (embedded in the lab) we worked on multidevice + sensor fusion (his expertise!) research.google/pubs/intent-dr… And eric j gonzalez built a whole tool for prototyping multidevice which we opensourced: github.com/google/xdtk A space where Andrea Colaço contributed a lot too!

We have two presentations @NeurIPS @UniReps workshop tomorrow: 1) Willow Han will present: openreview.net/forum?id=t4CnK…, and 2) Rouzbeh Meshkinnejad will present: openreview.net/forum?id=fS41j…

Say goodbye to perfect pinhole assumptions Excited to introduce 3DGUT—a Gaussian Splatting formulation that unlocks support for distorted cameras, including time dependent effects like rolling shutter, while maintaining the benefits of rasterization, rendering at >250 FPS. 🧵

Want to use Depth Anything, but need metric depth rather than relative depth? Thrilled to introduce Prompt Depth Anything, a new paradigm for accurate metric depth estimation with up to 4K resolution. 👉Key Message: Depth foundation models like DA have already internalized rich

New preprint! Category-biased patches encircle domain-general brain regions in the human lateral prefrontal cortex doi.org/10.1101/2025.0…

New paper out in Nature Human Behaviour! In it, Wilma Bainbridge and I find that participants tend to remember and forget the same speakers' voices, regardless of speech content. We also predict the memorability of voices from their low-level features: nature.com/articles/s4156…

We make about 3-4 fast eye movements a second, yet our world appears stable. How is this possible? In a preprint led by Luca Kämmer we test the intriguing idea that anticipatory signals in the fovea may explain visual stability. biorxiv.org/content/10.110…

Train your vision SAE on Monday, then again on Tuesday, and you'll find only about 30% of the learned concepts match. ⚓ We propose Archetypal SAE which anchors concepts in the real data’s convex hull, delivering stable and consistent dictionaries. arxiv.org/pdf/2502.12892…

I wrote a commentary on a very nice research paper that just appeared in Brain by Selma Lugtmeijer (she/her) Aleksandra Sobolewska Steven scholte. Spoiler: It's about modularity in mid-level vision. Here is the original paper: doi.org/10.1093/brain/… And here my commentary: doi.org/10.1093/brain/…

Next-gen vision pre-trained models shouldn’t be short-sighted. Humans can easily perceive 10K x 10K resolution. But today’s top vision models—like SigLIP and DINOv2—are still pre-trained at merely hundreds by hundreds of pixels, bottlenecking their real-world usage. Today, we

📷 Can AI understand camera motion like a cinematographer? Meet CameraBench: a large-scale, expert-annotated dataset for understanding camera motion geometry (e.g., trajectories) and semantics (e.g., scene contexts) in any video – films, games, drone shots, vlogs, etc. Links

Interesting work on memory: Memorability and sustained attention can be leveraged in real time to improve memory performance. #NEIfunded The University of Chicago Brady Roberts Wilma Bainbridge Monica Rosenberg Megan deBettencourt Springer Nature: bit.ly/3EoPrUA

I'm excited to announce publication of our new paper that can help answer age old questions of perspective in art history and #visionscience . nature.com/articles/s4159…

🔊New NVIDIA paper: Audio-SDS🔊 We repurpose Score Distillation Sampling (SDS) for audio, turning any pretrained audio diffusion model into a tool for diverse tasks, including source separation, impact synthesis & more. 🎧 Demos, audio examples, paper: research.nvidia.com/labs/toronto-a…

Excited to share our new #SIGGRAPH2025 paper! In this work, we show how to combine Gaussian splatting and computer-generated holography using Gaussian Wave Splatting. This enables photorealistic 3D holograms for emerging holographic VR/AR displays. 1/4

This week, we also have the honor of hosting Ida Momennejad as our guest speaker. Thank you, Ida Momennejad

Now in Nature! 🚀 Our method learns a controllable 3D model of any robot from vision, enabling single-camera closed-loop control at test time! This includes robots previously uncontrollable, soft, and bio-inspired, potentially lowering the barrier of entry to automation! Paper:

🚀New from Meta FAIR: today we’re introducing Seamless Interaction, a research project dedicated to modeling interpersonal dynamics. The project features a family of audiovisual behavioral models, developed in collaboration with Meta’s Codec Avatars lab + Core AI lab, that

Excited to announce that this paper is now out in Science Advances science.org/doi/10.1126/sc…

Our new study in Nature Computational Science, led by Haibao Wang, presents a neural code converter aligning brain activity across individuals & scanners without shared stimuli by minimizing content loss, paving the way for scalable decoding and cross-site data analysis. nature.com/articles/s4358…