iiinova — where your digital self keeps evolving

Anand Bhattad @anandbhattad.bsky.social · 08/09/2026

5/ None of the models we tested performed well on the Principia benchmark. Several interesting videos are on our website: principiabench.github.io Paper: arxiv.org/abs/2609.04200 PS: Principia is the latest step in a series of work from our group on exploring what generative models know (& don't).

principiabench.github.io

Principia: Relational Physics Tests for Video Models

Principia evaluates Newtonian physics in video models through calibration-independent relational consistency between paired objects.

010 0

Anand Bhattad @anandbhattad.bsky.social · 08/09/2026

4/ Using this relative consistency, we can completely circumvent the need for absolute measurements or other calibration. It provides an objective evaluation to measure models. We tested 6 different generators and 4 vision-language models on several hundred manually recorded and curated scenes.

100 0

Anand Bhattad @anandbhattad.bsky.social · 08/09/2026

3/ To address this, our approach shifts away from measuring single objects in isolation, focusing instead on comparing multiple objects to evaluate their relative consistency. For example, two blocks of different masses placed on identical inclines must reach the bottom at the same time.

100 0

Anand Bhattad @anandbhattad.bsky.social · 08/09/2026

2/ How do we verify physics accuracy in generated videos? The problem is that we cannot determine absolute time simply from the number of generated frames. Next, we do not know the spatial scale at which these models are operating, the camera’s coordinates, or the mass of the objects involved.

100 0

Anand Bhattad @anandbhattad.bsky.social · 08/09/2026

1/ What do video models know about high school physics? Less than you'd think. We built Principia: relational physics tests for video models. We evaluate on eight laws, over 500+ real scenes, and one simple idea. principiabench.github.io paper: arxiv.org/abs/2609.04200

100 0

Anand Bhattad @anandbhattad.bsky.social · 23/06/2026

📢 Super excited about our "Thinking in Boxes" Idea: fit 3D boxes around objects in a single "real" photo, then move the boxes. Objects translate & rotate; two of them can swap places; the camera moves; and the scene holds up. Enjoy the video below. More on our website: thinking-in-boxes.github.io

010 0

Anand Bhattad @anandbhattad.bsky.social · 02/06/2026

"Bitter Lessons" workshop tomorrow (Jun 3rd) starting from 08:45 am in Room 3A-3D at #CVPR2026. Website and schedule: sites.google.com/view/bitterl...

073 0

Reposted by Anand Bhattad

Vision and Graphics Trends @si-cv-graphics.bsky.social · 04/05/2026

𝗚𝗲𝗻𝗲𝗿𝗮𝗹𝗶𝘇𝗮𝗯𝗹𝗲 𝗦𝗽𝗮𝗿𝘀𝗲-𝗩𝗶𝗲𝘄 𝟯𝗗 𝗥𝗲𝗰𝗼𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻 𝗳𝗿𝗼𝗺 𝗨𝗻𝗰𝗼𝗻𝘀𝘁𝗿𝗮𝗶𝗻𝗲𝗱 𝗜𝗺𝗮𝗴𝗲𝘀 Vinayak Gupta, Chih-Hao Lin, Shenlong Wang ... Jia-Bin Huang arxiv.org/abs/2604.28193 Trending on www.scholar-inbox.com

011 0

Anand Bhattad @anandbhattad.bsky.social · 25/04/2026

... shift the blocks around and a pretrained diffusion model does the rendering. Blocks Worlds have been used in computer vision since the dawn of the field (Roberts' PhD thesis, 1963). We show how to make them work for something useful. Paper: arxiv.org/abs/2506.20703

001 0

Anand Bhattad @anandbhattad.bsky.social · 25/04/2026

At #ICLR2026, Vaibhav Vavilala presented our paper, "Generative Blocks World"! This is one of my favorites from recent times. We spent two years hacking around with diffusion models to get them to do object editing. The solution turned out to be way simpler. A few blocks in 3D space is enough ...

140 0

Reposted by Anand Bhattad

Zhenjun Zhao @ericzzj.bsky.social · 14/04/2026

SyncFix: Fixing 3D Reconstructions via Multi-View Synchronization Deming Li, Abhay Yadav, Cheng Peng, Rama Chellappa, @anandbhattad.bsky.social tl;dr: joint conditional over multiple views to enforce consistency during denoising arxiv.org/abs/2604.11797

041 0

Reposted by Anand Bhattad

Johns Hopkins Data Science and AI Institute @hopkinsdsai.bsky.social · 19/12/2025

Join us in advancing data science and AI research! The Johns Hopkins Data Science and AI Institute Postdoctoral Fellowship Program is now accepting applications for the 2026–2027 academic year. Apply now! Deadline: Jan 23, 2026. Details and apply: apply.interfolio.com/179059

0119 0

Anand Bhattad @anandbhattad.bsky.social · 06/10/2025

I decided not to travel to #ICCV2025 because it coincides with Diwali (Oct 20). Diwali often falls near the #CVPR deadline window, but this year overlaps with ICCV. I understand it’s hard to avoid all global holidays, but I hope future conferences can keep this in mind when selecting dates.

020 0

Anand Bhattad @anandbhattad.bsky.social · 06/10/2025

I will be recruiting a few students for Fall 2026. In particular, I will strongly consider a PhD applicant with training in applied/computational mechanics and computer vision/machine learning. If you or someone you know has this background, please contact me.

041 0

Anand Bhattad @anandbhattad.bsky.social · 21/09/2025

So You Want to Be an Academic? A couple of years into your PhD, but wondering: "Am I doing this right?" Most of the advice is aimed at graduating students. But there's far less for junior folks who are still finding their academic path. My candid takes: anandbhattad.github.io/blogs/jr_gra...

anandbhattad.github.io

So You Want to Be an Academic? What I Wish I Knew Early in Graduate School

Blog for junior PhD students on work, visibility, community, and sanity—long before the faculty job market is on the horizon.

1204 0

Anand Bhattad @anandbhattad.bsky.social · 30/06/2025

Thanks Andreas and the Scholar Inbox team! This is by far the best paper recommendation system I’ve come across. No more digging through overwhelming volumes and like the blog says, the right papers just show up in my inbox.

030 0

Reposted by Anand Bhattad

ML for Science @ml4science.bsky.social · 30/06/2025

On our blog: Science is moving fast. How do we keep up? #ScholarInbox, developed by the Autonomous Vision Group led by @andreasgeiger.bsky.social, helps researchers stay ahead - by making the discovery of #openaccess papers smarter and more personal: www.machinelearningforscience.de/en/scholar-i...

machinelearningforscience.de

Scholar Inbox: Daily Research Recommendations just for You

Science is moving fast. How can we keep up? Scholar Inbox helps researchers stay ahead by making the discovery of open access papers more personal.

12714 0

Anand Bhattad @anandbhattad.bsky.social · 30/06/2025

All slides from the #cvpr2025 (@cvprconference.bsky.social ) workshop "How to Stand Out in the Crowd?" are now available on our website: sites.google.com/view/standou...

000 0

Anand Bhattad @anandbhattad.bsky.social · 23/06/2025

This is probably one of the best talks and slides I have ever seen. I was lucky to see this live! Great talk again :)

130 0

Anand Bhattad @anandbhattad.bsky.social · 02/06/2025

A special shout-out to all the job-market candidates this year: it’s been tough with interviews canceled and hiring freezes🙏 After UIUC's blue and @tticconnect.bsky.social blue, I’m delighted to add another shade of blue to my journey at Hopkins @jhucompsci.bsky.social. Super excited!!

020 0

Anand Bhattad @anandbhattad.bsky.social · 02/06/2025

We will be recruiting PhD students, postdocs, and interns. Updates soon on my website: anandbhattad.github.io Also, feel free to chat with me @cvprconference.bsky.social #CVPR2025 I’m immensely grateful to my mentors, friends, colleagues, and family for their unwavering support.🙏

anandbhattad.github.io

Anand Bhattad - Research Assistant Professor

100 0

Anand Bhattad @anandbhattad.bsky.social · 02/06/2025

At JHU, I'll be starting a new lab: 3P Vision Group. The “3Ps” are Pixels, Perception & Physics. The lab will focus on 3 broad themes: 1) GLOW: Generative Learning Of Worlds 2) LUMA: Learning, Understanding, & Modeling of Appearances 3) PULSE: Physical Understanding and Learning of Scene Events

100 0

Anand Bhattad @anandbhattad.bsky.social · 02/06/2025

I’m thrilled to share that I will be joining Johns Hopkins University’s Department of Computer Science (@jhucompsci.bsky.social, @hopkinsdsai.bsky.social) as an Assistant Professor this fall.

182 0

Reposted by Anand Bhattad

Zhenjun Zhao @ericzzj.bsky.social · 08/05/2025

FastMap: Revisiting Dense and Scalable Structure from Motion Jiahao Li, Haochen Wang, @zubair-irshad.bsky.social, @ivasl.bsky.social, Matthew R. Walter, Vitor Campagnolo Guizilini, Greg Shakhnarovich tl;dr: replace BA with epipolar error+IRLS; fully PyTorch implementation arxiv.org/abs/2505.04612

0102 0

Anand Bhattad @anandbhattad.bsky.social · 29/04/2025

[2/2] However, if we treat 3D as a real task, such as building a usable environment, then these projective geometry details matter. It also ties nicely to Ross Girshick’s talk at our RetroCV CVPR workshop last year, which you highlighted.

010 0

Anand Bhattad @anandbhattad.bsky.social · 29/04/2025

[1/2] Thanks for the great talk and for sharing it online for those who couldn't attend 3DV. I liked your points on our "Shadows Don't Lie" paper. I agree that if the goal is simply to render 3D pixels, then subtle projective geometry errors that are imperceptible to humans are not a major concern.

110 0

Anand Bhattad @anandbhattad.bsky.social · 15/04/2025

Congratulations and welcome to TTIC! 🥳🎉

010 0

Anand Bhattad @anandbhattad.bsky.social · 02/04/2025

By “remove,” I meant masking the object and using inpainting to hallucinate what could be there instead.

000 0

Reposted by Anand Bhattad

Noah Snavely @snavely.bsky.social · 30/03/2025

This is really cool work!

181 0

Anand Bhattad @anandbhattad.bsky.social · 02/04/2025

Thanks Noah! Glad you liked it :)

000 0

Anand Bhattad @anandbhattad.bsky.social · 02/04/2025

[2/2] We also re-run the full pipeline *after each removal*. This matters: new objects can appear, occluded ones can become visible, etc, making the process adaptive and less ambiguous. Fig above shows a single pass. Once the top bowl is gone, the next "top" bowl gets its own diverse semantics too

000 0

Anand Bhattad @anandbhattad.bsky.social · 02/04/2025

[1/2] Not really... there's quite a bit of variation. When we remove the top bowl, we get diverse semantics: fruits, plants, and other objects that just happen to fit the shape. As we go down, it becomes less diverse: occasional flowers, new bowls in the middle, & finally just bowls at the bottom.

210 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[10/10] This project began while I was visiting Berkeley last summer. Huge thanks to Alyosha for the mentorship and to my amazing co-author Konpat Preechakul. We hope this inspires you to think differently about what it means to understand a scene. 🔗 visualjenga.github.io 📄 arxiv.org/abs/2503.21770

visualjenga.github.io

Visual Jenga: Discovering Object Dependencies via Counterfactual Inpainting

Visual Jenga is a new scene understanding task where the goal is to remove objects one by one from a single image while keeping the rest of the scene stable. We introduce a simple baseline that uses a...

010 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[9/10] Visual Jenga is a call to rethink what scene understanding should mean in 2025 and beyond. We’re just getting started. There’s still a long way to go before models understand scenes like humans do. Our task is a small, playful, and rigorous step in that direction.

100 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[8/10] This simple idea surprisingly scales to a wide range of scenes: from clean setups like a cat on a table or a stack of bowls... to messy, real-world scenes (yes, even Alyosha’s office).

210 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[7/10] Why does this work? Because generative models have internalized asymmetries in the visual world. Search for “cups” → You’ll almost always see a table. Search for “tables” → You rarely see cups. So: P(table | cup) ≫ P(cup | table) We exploit this asymmetry to guide counterfactual inpainting

120 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[6/10] We measure dependencies by masking each object, then using a large inpainting model to hallucinate what should be there. If the replacements are diverse, the object likely isn't critical. If it consistently reappears, like the table under the cat, it’s probably a support.

110 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[5/10] To solve Visual Jenga, we start with a surprising baseline without explicit physical reasoning & any 3D, simulation, or dynamics. Instead, we propose a training-free, generative approach that infers object removal order by exploiting statistical co-occurrence learned by generative models.

100 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[4/10] The goal of Visula Jenga is simple: 1) Remove one object at a time 2) Generate a sequence down to the background 3) Keep every intermediate scene physically & geometrically stable

100 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[3/10] Probing this understanding motivates our new task: Visual Jenga, a challenge beyond passive observation. Like in the game of Jenga, success demands understanding structural dependencies. Which objects can you remove without collapsing the scene? That’s where true understanding begins.

100 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[2/10] Today’s models can name everything in an image. But do they understand how a scene holds together? Inspired by Biederman’s classic work on scene perception + influential efforts by Hoiem et al, Bottou et al, & others, we ask: Can a model understand support structure and object dependencies?

110 0

Anand Bhattad @anandbhattad.bsky.social · 29/03/2025

[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

75914 0

Anand Bhattad @anandbhattad.bsky.social · 27/03/2025

I can’t believe this! Mind-blowing! There are small errors (a flipped logo, rotated chairs), but still, this is incredible!! Xiaoyan, who’s been working with me on relighting, sent this over. It’s one of the hardest examples we’ve consistently used to stress-test LumiNet: luminet-relight.github.io

030 0

Reposted by Anand Bhattad

Jia-Bin Huang @jbhuang0604.bsky.social · 15/03/2025

Check out UrbanIR - Inverse rendering of unbounded scenes from a single video! It’s a super cool project led by the amazing Chih-Hao! @chih-hao.bsky.social is a rising star in 3DV! Follow him! Learn more here👇

0102 0

Anand Bhattad @anandbhattad.bsky.social · 16/03/2025

Can we create realistic renderings of urban scenes from a single video while enabling controllable editing: relighting, object compositing, and nighttime simulation? Check out our #3DV2025 UrbanIR paper, led by @chih-hao.bsky.social that does exactly this. 🔗: urbaninverserendering.github.io

021 0

Reposted by Anand Bhattad

Johan Edstedt @parskatt.bsky.social · 11/03/2025

Introducing DaD (arxiv.org/abs/2503.07347), a pretty cool keypoint detector. As this will get pretty long, this will be two threads. The first will go into the RL part, and the second on the emergence and distillation.

46211 0

Reposted by Anand Bhattad

Ben Waber @bwaber.bsky.social · 06/03/2025

Last was an intriguing talk by @anandbhattad.bsky.social on the degree to which physical properties are encoded in generative image models at the @grasplab.bsky.social www.youtube.com/watch?v=Idcb... (7/7)

youtube.com

Fall 2024 GRASP Seminar Anand Bhattad, Toyota Technological Institute at Chicago

YouTube video by GRASP Lab

011 0

Anand Bhattad @anandbhattad.bsky.social · 28/02/2025

Exciting work led by Zitian & @freudfc.bsky.social. This is their first Computer Vision paper acceptance! 🚀 w/ Mathieu Garon and @jflalonde.bsky.social

031 0

Anand Bhattad @anandbhattad.bsky.social · 28/02/2025

ZeroComp will be presented as an Oral at #WACV2025! Train a diffusion model as a renderer that takes intrinsic images as input. Once trained, we can perform zero-shot object compositing & can easily extend this to other object editing tasks, such as material swapping. arxiv.org/abs/2410.08168

140 0