Roy Velich (@RoyVelich) profile picture

Roy Velich

@RoyVelich

Phd Candidate, Technion Geometry and Deep Learning

 Haifa, Israel
Joined September 2022
  • Tweets523
  • Following402
  • Followers43
6Media
Photo by Roy Velich (@RoyVelich)Photo by Roy Velich (@RoyVelich)Photo by Roy Velich (@RoyVelich)Photo by Roy Velich (@RoyVelich)Photo by Roy Velich (@RoyVelich)
We're giving away a Hogwarts Legacy custom wrapped NVIDIA GeForce RTX 5090 to celebrate its official release with DLSS 5 ๐Ÿช„ To enter: ๐ŸŸข Share this post ๐ŸŸข Comment #RTXON Terms and Conditions: nvidia.com/en-us/geforce/conโ€ฆ
Image from NVIDIA GeForce (@NVIDIAGeForce)
84,991
72,424
7,717
67,803
5,942,481
Roy Velich retweeted
Noam Rotstein (@NoamRot) profile picture
Noam Rotstein
@NoamRotNov 2
Meet ReLift, our latest work on 3D tracking! ๐Ÿ‹๏ธ Rethinking 3D tracking with the right combination: Strong 2D trackers ๏ผ‹ 3D geometry ๏ผ‹ Efficient trajectory corrector โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€ ใ€€ Dense 3D tracking, ใ€€ even through occlusions!
Assaf Singer (@assaf_singer) profile pictureAssaf Singer
@assaf_singerNov 2
๐Ÿš€ Introducing ReLift: dense 3D tracking from monocular video. Strong 2D trackers provide correspondences. Geometry models produce depth & cameras. We show how to combine them, repair 3D errors, and recover occluded motion for SOTA 3D tracking. All it takes: one small corrector + off-the-shelf models. ๐ŸŒ Project: relift3d.github.io/ @MarounKhriesh @NoamRot @yoni_kasten @orlitany
1
3
31
1,723
Roy Velich retweeted
Noam Rotstein (@NoamRot) profile picture
Noam Rotstein
@NoamRotOct 21
๐ŸŽฌ Excited to share our paper: Joints, Camera, Action! (JCA) JCA generates video & 3D motion together for any rig! Video models generate motion you can watch. But what if you could use that motion to animate a 3D asset, with motion that responds to its surroundings? research.nvidia.com/labs/parโ€ฆ ๐Ÿงต
4
42
6
99
7,341
Happy to share that our paper has been accepted to #CVPR2026! Learning Eigenstructures of Unstructured Data Manifolds๐Ÿ“„ arxiv.org/abs/2512.01103 We learn spectral bases directly from unstructured point clouds โ€” no operator construction, no meshing, no eigensolver. Just a network that recovers eigenfunctions, eigenvalues, and the implicit sampling density from scratch. Big thanks to my co-authors Arkadi Piven, David Bensaรฏd, @tom_dages, Daniel Cremers, Ron Kimmel. See you in Denver!
Learning Eigenstructures of Unstructured Data Manifolds

Learning Eigenstructures of Unstructured Data Manifolds

We introduce a novel framework that directly learns a spectral basis for shape and manifold analysis from unstructured data, eliminating the need for traditional operator selection,...

arxiv.org
1
4
580
Shortest paths change with missing parts. To focus on consistent geodesic pairs, we created the Wormhole criterion, combining intrinsic and extrinsic geometry, for shape analysis and shape matching. Wormhole Loss for Partial Shape Matching arxiv.org/abs/2410.22899 #NeurIPS2024
Image from Thomas Dagรจs (@tom_dages)
2
11
538
ื‘ื ืชื™ื™ื, ื‘ืงื‘ื•ืฆืช ื”ื’ื™ืื•-ืžื™ืžื– ืฉืœ ืžื•ืขื“ื•ืŸ ื”ืžืฉื—ืง ื”ื’ื“ื•ืœ
14
39
2
649
31,360
๐Ÿšจ๐Ÿšจ๐Ÿšจ New paper alert! ๐—ฃ๐—ฎ๐˜๐—ต๐˜„๐—ฎ๐˜†๐˜€ ๐—ผ๐—ป ๐˜๐—ต๐—ฒ ๐—œ๐—บ๐—ฎ๐—ด๐—ฒ ๐— ๐—ฎ๐—ป๐—ถ๐—ณ๐—ผ๐—น๐—ฑ: ๐—œ๐—บ๐—ฎ๐—ด๐—ฒ ๐—˜๐—ฑ๐—ถ๐˜๐—ถ๐—ป๐—ด ๐˜ƒ๐—ถ๐—ฎ ๐—ฉ๐—ถ๐—ฑ๐—ฒ๐—ผ ๐—š๐—ฒ๐—ป๐—ฒ๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป We propose a new approach to image editing! ๐Ÿ“„huggingface.co/papers/2411.1โ€ฆ ๐Ÿ”— arxiv.org/abs/2411.16819 [1/8]
Image from Noam Rotstein (@NoamRot)
3
24
2
90
13,667
Dear @hammockmusic , Andalusia is one of my all-time favorites, a truly magical track. I always wondered about its name though. Is it named after the Spanish region? If so, what's the background story? ๐Ÿ˜… It always baffled me. Thanks!
2
3
358
Load more