Or Litany (@orlitany)

Or Litany reposted

Elisabetta Fedele @efedele16

Jul 3

The submissions to our @eccvconf workshop OpenSUN3D on Open-World 3D Scene Understanding and Representations are open! 🇸🇪✨🤖 🗓️ Deadlines are on the 1st of August for paper tracks and on the 25th of August for challenges! 🌍: opensun3d.github.io 📝: openreview.net/group?id=thec…

2

4

15

1,191

Or Litany @orlitany

Jun 29

Fantastic work -- principled and works great. Congrats @OrPerel and team!

Or Perel @OrPerel

Jun 29

🎉 Excited to introduce TRON, a relighting framework for 3D captures. 💡TRON pairs a neural renderer with 3D Gaussian reconstructions, achieving realistic quality, with 3D, material, & lighting control at interactive frame rates. arxiv.org/abs/2606.11314 research.nvidia.com/labs/sil…

3:36

1

11

2,478

Or Litany reposted

International Conference on 3D Vision (3DV) @3DVconf

Jun 20

Got unlucky with #ECCV2026 reviews? Strong #3D #research deserves a good home! 👉 Integrate feedback, make your paper bulletproof & submit to #3DV2027! 💪 🎯 Focused audience on #3D topics ⏳ Deadline: August 28 📍 Thessaloniki, Greece 👉 Spread the word! 3dvconf.github.io

7

43

4,229

Or Litany @orlitany

Jun 20

SpectralSplats is accepted at #ECCV2026! 🎉 Tracking 3DGS across frames is harder than it looks — appearance losses break the moment the pose drifts. Spectral moments keep it robust. @eccvconf

Or Litany @orlitany

Mar 26

1/9 Excited to share SpectralSplats! 📢 Given a 3DGS asset target video, we deform it to match the video via differentiable rendering. Appearance-based tracking fails when the initial pose is even slightly off. Our spectral loss stays robust. 🔗 avigailco.github.io/Spectral… 🧵

0:04

1

5

48

4,098

Or Litany @orlitany

Jun 20

Excited that RadarGen is accepted to #ECCV2026! Looking forward to chatting radar simulation in Malmö — let’s push this underexplored topic forward. @eccvconf

Or Litany @orlitany

22 Dec 2025

🚗📡Radar is the unsung hero of AV perception: widespread in cars, yet overlooked in simulation. Introducing RadarGen: Realistic radar synthesis from cameras using diffusion. Massive kudos to my fantastic team at @TechnionLive and @NVIDIAAI radargen.github.io/

1

4

39

3,944

Or Litany @orlitany

Jun 19

📢 New paper: FlowBender. Conditional generators drift from their own conditioning. The usual fix: tune guidance & pray 🙏 💡We train them to self-correct from their own error 4.5 dB on 3D texturing, 4.9 dB on SR. Give it a spin and bend away👇 flow-bender.github.io

Daniel Gilo @danielgilo1

Jun 19

Conditional diffusion/flow models often produce outputs inconsistent with the very signal conditioning them. The error is easily measurable, yet models are never trained to act on it. In FlowBender (now on arXiv), we train the model to correct its own errors. 🧵

0:08

1

12

57

8,819

Or Litany @orlitany

Jun 12

[7/7] Kudos to @mann_amir_ for leading this project, and to fantastic teammates @galhar6 and Merav Keidar. We welcome any feedback! #ComputerVision #HumanMotion #Diffusion

1

192

Or Litany @orlitany

Jun 12

[6/7] On fitness videos (Fit3D), our method achieves 2× lower joint error than WHAM, and 5.5× smoother motion. On MAS's NBA dataset, VideoMDM is preferred by humans 64% of the time over MAS.

0:07

1

217

Or Litany @orlitany

Jun 12

[5/7] On a synthetic, 2D-only version of HumanML3D, VideoMDM achieves FID 0.88 -- nearly closing the gap to fully 3D-supervised MDM (0.54), and ~2× better than the strongest 2D-supervised baseline.

0:03

1

136

Or Litany @orlitany

Jun 12

[4/7] 3D motion representations add channels derived from joint positions -- rotations, velocities, foot contacts -- with no direct 2D supervision. Our fix: snap predicted joints to the rays induced by 2D points, then recompute them, providing pseudo-targets for every channel.

1

213

Or Litany @orlitany

Jun 12

[3/7] We suggest a depth-weighting loss to cancel the perspective bias, and prove this depth-weighted 2D loss is, in expectation, equivalent to direct 3D supervision.

1

325

Or Litany @orlitany

Jun 12

The challenge: we want a 3D denoiser, but can only supervise in 2D. Building on our ICCV'25 "Lesson in Splats", we use a crude 2D-to-3D lifter as teacher -- once noised, it is close enough to true 3D noise to train on. The denoiser works in 3D; supervision via 2D reprojection.

1

491

Or Litany @orlitany

Jun 12

Excited to share our new paper: VideoMDM 📢 We propose a principled framework for training 3D motion diffusion models (e.g. MDM), using only 2D supervision from monocular videos -- no 3D ground truth required. Project: videomdm.github.io Paper: arxiv.org/abs/2606.13364 🧵

0:33

2

21

130

9,226

Or Litany reposted

Technion Israel

@TechnionLive

Jun 11

A new technology developed at the Technion could significantly expand access to AI-powered video creation Called Time to Move (TTM), the system enables users to create and precisely control realistic AI-generated video clips using simple mouse movements. Unlike previous approaches, TTM requires no retraining, no additional computational cost, and can be integrated directly into existing video-generation models. Developed by Dr. @orlitany , Prof. Ron Kimmel, and students Asaf Singer, Noam Rotstein, and Amir Mann from the Henry and Marilyn Taub Faculty of Computer Science, TTM was presented at @iclr_conf 2026, one of the world's premier conferences in deep learning and artificial intelligence.

2

6

31

2,956

Or Litany @orlitany

Jun 7

This morning @CVPR: [poster #35] we explain how to — consistent edits of multiview Images without training @danielgilo1 and I will be there to walk you through the method

Or Litany @orlitany

Feb 25

Excited to share our new paper, accepted to CVPR 2026: Instruct-Mix2Mix! 📢 We tackle the challenge of sparse multi-view editing: modifying a scene given only a few images (e.g., 4-8) via text instructions. The code is available now. Project page: danielgilo.github.io/instruc… 🧵

2

14

1,565

Or Litany @orlitany

Jun 6

This afternoon @CVPR: Realiz3D [poster #526] Come learn how to make image diffusion models 3D aware without leaking synthetic appearance.

Ido Sobol @IdoSobol

Apr 3

Tired of 3D generations that look synthetic? They don’t have to. Excited to share that Realiz3D is accepted to #CVPR2026 🎉 A framework for training diffusion models that are 3D-consistent, controllable, and photorealistic. 🔗 Project page: idosobol.github.io/realiz3d/ 🎥 Teaser ↓

5

26

2,001

Or Litany reposted

Francis Engelmann ☀️ @CVPR'26

@FrancisEngelman

Jun 6

Today afternoon @CVPR: 𝐅𝐮𝐧𝐑𝐞𝐜 [Poster 654] 🌴 ➡️Input: Egocentric interaction Video 👷‍♀️ ⬅️Output: Functional 4D Reconstruction 🏡 -- come say hi at the poster! 🚀 /w @AlexDelitzas @mapo1 @majti89 @Stanford @ETH

0:15

Alex Delitzas

@AlexDelitzas

Jun 6

1/ Excited to present FunREC: "Reconstructing Functional 3D Scenes from Egocentric Interaction Videos" at #CVPR2026! FunREC builds functional, interactable 3D digital replicas of real-world environments from a single egocentric interaction video. 👇 Here's me turning my own kitchen into a functional 3D digital twin, just by interacting with it. 🔗 Project page: functionalscenes.github.io 📄 Paper: arxiv.org/abs/2604.05621 🗓️ Poster: Session 4 (#654), Sat Jun 6, 16:45–18:45, ExHall A

1:00

2

16

128

15,557

Or Litany reposted

Angela Dai @angelaqdai

Jun 2

Come by our ScanNet workshop @CVPR June 3 in 710, 1:00pm onwards! 5 exciting keynotes on world models, NVS, 3D gen, understanding and more, from @taiyasaki @davnov134 @orlitany @PeterHedman3 @RamananDeva and talks from NVS semantics benchmark winners: scannetpp.mlsg.cit.tum.de/sc…

22

69

19,199

Or Litany reposted

International Conference on 3D Vision (3DV) @3DVconf

May 25

Looking for a nice home for your paper? #3DV2027 is waiting! 🎉 ⏰ Conference dates: April 6-9, 2027 ✈️ Place: #Thessaloniki (#SKG), #Greece 🇬🇷 📝 Paper ddl: Aug 28 🎥 Supp ddl: Sep 02 🆕 Rebuttal only upon invite for borderline papers! #CallForPapers: 3dvconf.github.io/2027/call-…

3

22

93

10,963

Or Litany @orlitany

May 15

Excited to finally share that our paper is now available on arXiv and Hugging Face 🎉 arxiv.org/abs/2605.13852 Looking forward to presenting it at CVPR 2026 in Denver this June!

Realiz3D: 3D Generation Made Photorealistic via Domain-Aware Learning

We often aim to generate images that are both photorealistic and 3D-consistent, adhering to precise geometry, material, and viewpoint controls. Typically, this is achieved by fine-tuning an image...

arxiv.org

Ido Sobol @IdoSobol

Apr 3

Tired of 3D generations that look synthetic? They don’t have to. Excited to share that Realiz3D is accepted to #CVPR2026 🎉 A framework for training diffusion models that are 3D-consistent, controllable, and photorealistic. 🔗 Project page: idosobol.github.io/realiz3d/ 🎥 Teaser ↓

6

36

5,015