I found this old bassoon sheet music in storage, and tried to have Gemini scan it to PDF. I think it is getting better with music? But still goes off the rails. (ChatGPT does even worse.)

The CVPR process seems driven by the goal of extracting 3 reviews for each paper, a goal that seems to lead to a lot of angst. Is there a reason why 3 is a magic number? Why not two, or even one (assuming the quality is high)?

If you are looking for a new podcast to get into, I recommend Topics with Michael Ian Black and Michael Showalter. Even though they are comedians, they aren't afraid to dive into some weighty topics! www.earwolf.com/show_archive...

What are some examples of computer vision papers that have attractive system diagrams?

Over the past year, my lab has been working on fleshing out theory + applications of the Platonic Representation Hypothesis. Today I want to share two new works on this topic: Eliciting higher alignment: arxiv.org/abs/2510.02425 Unpaired learning of unified reps: arxiv.org/abs/2510.08492 1/9

#TTT3R: 3D Reconstruction as Test-Time Training TTT3R offers a simple state update rule to enhance length generalization for #CUT3R — No fine-tuning required! 🔗Page: rover-xingyu.github.io/TTT3R We rebuilt @taylorswift13’s "22" live at the 2013 Billboard Music Awards - in 3D!

We present a new approach to inference-time scene optimization, which we name Radiant Triangle Soup (RTS) www.arxiv.org/abs/2505.23642. Also check out really great concurrent work from Held et al. @janheld.bsky.social, Triangle Splatting arxiv.org/abs/2505.19175

🧠How “old” is your model? Put it to the test with the KiVA Challenge: a new benchmark for abstract visual reasoning, grounded in real developmental data from children and adults. 🏆 Prizes: 🥇$1K to the top model 🥈🥉$500 📅 Deadline: 10/7/25 🔗 kiva-challenge.github.io @iccv.bsky.social

ChatGPT and Gemini both seem to struggle with sheet music. They both insist that this excerpt is in D major (2 sharps), and resist any attempt to tell them that there 3 sharps in the key signature. I think this is really cool and interesting!

Think LMMs can reason like a 3-year-old? Think again! Our Kid-Inspired Visual Analogies benchmark reveals where young children still win: ey242.github.io/kiva.github.... Catch our #ICLR2025 poster today to see where models still fall short! Thurs. April 24 3-5:30 pm Halls 3 + 2B #312

Dynamic Camera Poses and Where to Find Them Chris Rockwell, @jtung.bsky.social, Tsung-Yi Lin, Ming-Yu Liu, David F. Fouhey, Chen-Hsuan Lin tl;dr: a large-scale dataset of dynamic Internet videos annotated with camera poses arxiv.org/abs/2504.17788

1/6 🔍➡️ How to transform standard videos into immersive 360° panoramas? We've designed a new AI system for video-to-360° panorama generation! Our key insight: large-scale data is crucial for robust panoramic synthesis across diverse scenes.

We have released the Stereo4D dataset! Explore the real-world dynamic 3D tracks: github.com/Stereo4d/ste...

This is really nice work on visual discovery from @boyangdeng.bsky.social!

We're very excited to introduce TAPNext: a model that sets a new state-of-art for Tracking Any Point in videos, by formulating the task as Next Token Prediction. For more, see: tap-next.github.io

A thread of thoughts on radiance fields, from my keynote at 3DV: Radiance fields have had 3 distinct generations. First was NeRF: just posenc and a tiny MLP. This was slow to train but worked really well, and it was unusually compressed --- The NeRF was smaller than the images.

Fifth Ave jammed #handsoff

🚀 We’ve just released the code and checkpoints for our #ICLR2025 Oral paper: "LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias". Check it out below 👇 🔗 Code: github.com/haian-jin/LVSM 📄 Paper: arxiv.org/abs/2410.17242 🌐 Project Page: haian-jin.github.io/projects/LVSM/

This is really cool work!

[1/10] Is scene understanding solved? Models today can label pixels and detect objects with high accuracy. But does that mean they truly understand scenes? Super excited to share our new paper and a new task in computer vision: Visual Jenga! 📄 arxiv.org/abs/2503.21770 🔗 visualjenga.github.io

#Backslash at #CornellTech, dedicated to advancing new works of art and technology that escape convention, has announced Mimi Ọnụọha as its first Backslash Fellow: tech.cornell.edu/news/mimi-on... “This work feels like a marked evolution for me personally,” said Ọnụọha. @snavely.bsky.social

#CVPR2025 offers registration and travel support to students from underrepresented communities. Awards are based on need, contribution, travel distance, identity, and advisor support. Information and form: forms.gle/uDR2Q74drC4V...

Exciting news! MegaSAM code is out🔥 & the updated Shape of Motion results with MegaSAM are really impressive! A year ago I didn't think we could make any progress on these videos: shape-of-motion.github.io/results.html Huge congrats to everyone involved and the community 🎉

I think Qianqian et al's work is really cool! The problem of modeling state within a 3D reasoning system is quite interesting. (And I believe it's pronounced "cuter".)

Late to post, but excited to introduce CUT3R! An online 3D reasoning framework for many 3D tasks directly from just RGB. For static or dynamic scenes. Video or image collections, all in one! Project Page: cut3r.github.io Code and Model: github.com/CUT3R/CUT3R

This is really unimportant, but I keep seeing the word "advancements" in writing where I would have seen the word "advances" before. I'm taking this to mean that LLMs are at play and therefore, they will influence the language such that the two words will eventually come to mean the same thing!

This is a really fun podcast episode for people whose favorite part of the Empire Strikes Back film score is when the Millenium Falcon flies out of the giant space slug! www.settlingthescorepodcast.com/66-the-empir... @scoresettlers.bsky.social

My question for the history heads out there: In the 1960s, did they call the 1880s "the Eighties"?

Announcing Diffusion Forcing Transformer (DFoT), our new video diffusion algorithm that generates ultra-long videos of 800+ frames. DFoT enables History Guidance, a simple add-on to any existing video diffusion models for a quality boost. Website: boyuan.space/history-guidance (1/7)

A shoutout to KWAX-FM, public classical radio broadcasting from Eugene, Oregon. I think it is a real gem! They do Listener's Choice Friday, where you can call in and make a request (they've taken my calls even though I'm in NYC). kwax.uoregon.edu

New paper about pictures: I identify trends in geometric perspective in my own drawings and photos, and compare them to how the original scenes looked. I discuss what these trends might say about art history and vision science. Published in _Art & Perception_. #visionscience psyarxiv.com/pq8nb

Did Italo Calvino discover bag of words and topic models in 1979?

"𝐑𝐚𝐝𝐢𝐚𝐧𝐭 𝐅𝐨𝐚𝐦: Real-Time Differentiable Ray Tracing" A mesh-based 3D represention for training radiance fields from collections of images. radfoam.github.io arxiv.org/abs/2502.01157 Project co-lead by my PhD students Shrisudhan Govindarajan and Daniel Rebain, and w/ co-advisor Kwang Moo Yi

Continuous 3D Perception Model with Persistent State Qianqian Wang, Yifei Zhang, @holynski.bsky.social, Alexei A. Efros, Angjoo Kanazawa tl;dr: DUSt3R+visual tokens from input image via ViT encoder+current/past tokens interaction via ViT decoders; dynamic&unobserved arxiv.org/abs/2501.12387

Sharing this really exciting job opportunity I came across!

This vintage ad for Braniff International strikes me as so bizarre and off-kilter, almost as if it was AI generated. "Space helmets"?? youtu.be/AcyVkRXq7QI?...

Nice view from NYC of the International Space Station passing very close to Jupiter just now. I like these webpages, which make it easy to know when to check outside for the ISS! james.darpinian.com/satellites/ spotthestation.nasa.gov/signup.cfm

New paper! A self-supervised object-centric 2.1D image representation using 3D Gaussians, extending MAE with a Gaussian bottleneck. While Gaussian splatting has been used for single-scene reconstruction, we’re the first to apply it to image representation learning! brjathu.github.io/gmae/.

Image matching and ChatGPT - new post in the wide baseline stereo blog. tl;dr: it is good, even feels like human, but not perfect. ducha-aiki.github.io/wide-baselin...

🤔Can Generative Video Models Help Pose Estimation? ✅Yes! We find that generative video models can hallucinate plausible intermediate frames that provide useful context for pose estimators (e.g. DUSt3R), especially for images with little to no overlap. 🔗 inter-pose.github.io

On our holiday nostalgia subway ride from 96th St, we were treated to a live announcement from Mr. Charlie "Stand Clear of the Closing Doors, Please" Pellett.

This is great! Check out more about @linyijin.bsky.social's wonderful work towards a dynamic DUSt3R at stereo4d.github.io!

Introducing 👀Stereo4D👀 A method for mining 4D from internet stereo videos. It enables large-scale, high-quality, dynamic, *metric* 3D reconstructions, with camera poses and long-term 3D motion trajectories. We used Stereo4D to make a dataset of over 100k real-world 4D scenes.

Introducing Doppelgangers++! 🚀 An enhanced pairwise image classifier that tackles visual aliasing (doppelgangers) to improve 3D reconstruction accuracy across diverse, real-world scenes. 🌍✨ 🔗Project page: bit.ly/3VAPMJc. Code is also available.

Introducing MegaSaM! Accurate, fast, & robust structure + camera estimation from casual monocular videos of dynamic scenes! MegaSaM outputs camera parameters and consistent video depth, scaling to long videos with unconstrained camera paths and complex scene dynamics!

Zhengqi led a really nice new paper that computes really nice camera poses and depth maps from everyday videos, not just the standard SLAM-mable sorts of videos you often see. It's really fast and robust, and I think it's quite neat.

💡Introducing ScribbleLight, a generative model that can relight an image with simple scribbles - making it easier for virtual staging and interior design of homes. Led by Jun-Myeong, collabs with Anand (TTIC), Pieter (W&M), and Annie. @unccs.bsky.social 👉 chedgekorea.github.io/ScribbleLight/

Check out CAT4D: our new paper that turns (text, sparse images, videos) => (dynamic 3D scenes)! I can't get over how cool the interactive demo is. Try it out for yourself on the project page: cat-4d.github.io

I'm building a new research lab @ Uni of Cambridge focusing on 4D computer vision and generative models. Interested in joining us as a PhD student? Apply to the Engineering program by Dec 3 🗓️ www.postgraduate.study.cam.ac.uk/courses/dire... ChatGPT's "portrait of my current life"👇 elliottwu.com

This has been fun! I really enjoyed A'ing these cool Q's. I am still trying to get through more of them (but it is taking me a long time, because I'm really slow at thinking).