@Jimantha

3D vision fanatic. Professor @cornell_tech & Researcher @GoogleDeepmind. He or they. https://xitter.cf/t.co/m7Rs5xUFfG

 New York, NY
snavely.io
Joined June 2008
  • Tweets928
  • Following870
  • Followers10K
56Media
Lucky to have a signed copy!
Paul-Edouard Sarlin@pesarlinSep 10
Honored and humbled to receive the ECVA PhD Award at #ECCV2026 @eccvconf! So grateful to my advisor @mapo1 for his guidance throughout, to all my brilliant coauthors over the years, and to the computer vision community for the support and the journey together. Thank you all
1
1
60
6,028
Noah Snavely retweeted
Sara Sabour
@sabour_saraAug 27
🚀 We are excited to release the dataset, code, and expanded paper for ORBIT++: Benchmarking SfM in the Wild with 360° Video! Following our initial publication at CVPR, we’re releasing the full benchmark and codebase: 🌐 Project: orbit-sfm.github.io/ 📄 arXiv: arxiv.org/abs/2608.22039 💻 Code/Data: github.com/google-deepmind/o… (1/3)
2
20
1
91
7,337
Noah Snavely retweeted
Shiry Ginosar@shiryginosarJul 22
Introducing FilmGPT 🎬 (#SIGGRAPH2026) Can we learn the grammar of film directly from movies? We train an AR transformer on feature films to edit raw footage into coherent cinematic sequences, without generating new pixels. Project page: mudtriangle.com/filmgpt Demo below👇
4
12
2
92
8,013
Noah Snavely retweeted
Baback Elmieh
@BabackEJul 22
Picture stepping into a live 3D broadcast or 3D video call and looking around freely. That is the promise of dynamic NVS. But multi-view camera setups are fixed so parts stay hidden. Looking anywhere means recovering them from memory, which current stateful and stateless models cannot do in real-time over long-horizons. Until now. Introducing NSTM. ⚡️Real-time inference + 🧠minute-long memory.
7
12
2
35
4,334
We use a sort of @elliottszwu-style strategy to build a dataset of symmetries in architectural scenes, then detect and localize symmetries in new images via signed-distance functions. This is really great work from Hanyu with very nice visuals!
Hanyu Chen@hanyuc1110Jun 5
Excited to share ArchSym at #CVPR2026! 🏛️ Existing 3D symmetry detectors work well on clean, object-centric data. But what about in the wild? In our work, we tackle 3D-grounded reflectional symmetry detection specifically for real-world architectural landmarks. 🧵[1/7]
2
5
39
5,884
Thank you, Humphrey! It's a real honor to receive this award!
Humphrey Shi
@humphrey_shiJun 5
Congratulations to Noah Snavely @Jimantha from @Cornell receiving the 2026 Thomas S. Huang Memorial Prize at @CVPR — a well-deserved recognition of his impactful contributions to computer vision research and our community. Remembering Tom & Margaret: youtu.be/QV7WnO9Lk9M?si=iYmA…
27
6
1
186
25,663
Noah Snavely retweeted
Hanyu Chen@hanyuc1110Jun 5
Excited to share ArchSym at #CVPR2026! 🏛️ Existing 3D symmetry detectors work well on clean, object-centric data. But what about in the wild? In our work, we tackle 3D-grounded reflectional symmetry detection specifically for real-world architectural landmarks. 🧵[1/7]
1
6
1
23
7,980
Noah Snavely retweeted
Yuanbo Xiangli@ambie_kkJun 5
Honey, I Shrunk the Arc de Triomphe! 😱 Ever notice how SOTA depth models suffer from "scale-collapse"—metrically shrinking distant landmarks like they're toys? We introduce MetricScenes: a new in-the-wild metric dataset that fixes this!
2
18
3
157
23,430
A really nice paper with a really cool title from Yuanbo, Hanyu, and Xueqing!
Yuanbo Xiangli@ambie_kkJun 5
Honey, I Shrunk the Arc de Triomphe! 😱 Ever notice how SOTA depth models suffer from "scale-collapse"—metrically shrinking distant landmarks like they're toys? We introduce MetricScenes: a new in-the-wild metric dataset that fixes this!
2
27
4,425
Noah Snavely retweeted
Haian Jin
@Haian_JinJun 3
This work has been fully open-sourced, including all code, model weights, and data related scripts. Feel free to check it out! Our CVPR poster session will be on Saturday, June 6, from 4:45 to 6:45 PM in ExHall A & F 2. Our team (@holynski_, @jon_barron , @ChrisWu6080, @Jimantha, and me) would be onsite and very happy to discuss it in person!
Haian Jin
@Haian_JinMar 5
Spatial reconstruction is a long-context problem: real scenes come with hundreds of images. But O(N²) transformer-based models don’t scale efficiently. Introducing: 🤐ZipMap (CVPR ’26): Linear-Time, Stateful 3D Reconstruction via Test-Time Training (TTT). ZipMap “zips” a large image collection into an implicit TTT scene state in a single linear-time operation. The state will then be decoded into spatial outputs, and can be queried efficiently for novel-view geometry and appearance (~100 FPS) ZipMap is not only much faster (>20× faster than VGGT), but also matches or surpasses the accuracy of all SOTA models.
1
8
2
51
10,698
Noah Snavely retweeted
Aditya Chetan
@justachetanMay 29
Humans can watch tasks like cooking or assembly and reason about what happened, when, and between which parts. Can LVLMs do the same? We built Flat-Pack Bench to test this – and found there is still a long way to go. Accepted at #CVPR2026! 🎥🪑🧩(1/n)
1
12
2
40
10,032
Noah Snavely retweeted
Bharath Raj@bharathrajn98May 28
Feed-forward 3D reconstruction methods typically predict pointmaps in camera-centric frames. But why should a camera's arbitrary orientation define the coordinate system? We introduce G3T, a transformer that predicts pointmaps in gravity-aligned frames. Regardless of input image orientation, our method always produces upright pointmaps (see demo). We leverage this uprightness to create G3T-Long, a submap-based reconstruction method that improves robustness on long-sequence 3D reconstruction (more on that below). Interactive demos, code, and model weights are available on our project page.
8
52
2
307
24,680
Noah Snavely retweeted
Linyi Jin@jin_linyiMay 20
Excited to share what we are building -- Genie experience grounded in real-world street view. Try it out at labs.google/fx/projectgenie
🤖 Made with AI
Google
@GoogleMay 19
Project Genie is a @GoogleLabs experiment that lets you simulate dynamic worlds you can navigate in real time with Genie, our general-purpose world model. Today, we’re connecting Project Genie to nearly 20 years of Street View data from Google Maps — so you can now build interactive spaces based on real-world locations. Street View imagery in Project Genie is available now for places in the U.S., and will expand to more locales over time. #GoogleIO
8
42
4
424
68,011
Noah Snavely retweeted
Ruojin Cai@ruojin8May 3
Excited to share our CVPR’26 work! Moving beyond dense captures to long-tail Internet photos, we introduce MegaDepth-X and sparsity-aware sampling for 3D reconstruction. Great work led by Yuan (@yuanli16342871) and the team. Project page: megadepth-x.github.io
Yuan Li@yuanli16342871May 3
Introducing: Long-Tail Internet Photo Reconstruction (CVPR’26) We go beyond densely captured imagery to train more general 3D foundation models for the long tail of noisy, sparse, incomplete Internet photo collections of 3D scenes. Yet, we face a data bottleneck: models need ground truth for these long-tail scenes, which classical SfM fails to provide. How do we bypass it? We break this bottleneck with two key contributions. First, we introduce MegaDepth-X, a large new dataset of scenes with high-quality 3D supervision. Second, we propose a new way to simulate difficult image sets for training. Project page: megadepth-x.github.io/
12
1
110
13,300
Noah Snavely retweeted
Yuan Li@yuanli16342871May 3
Introducing: Long-Tail Internet Photo Reconstruction (CVPR’26) We go beyond densely captured imagery to train more general 3D foundation models for the long tail of noisy, sparse, incomplete Internet photo collections of 3D scenes. Yet, we face a data bottleneck: models need ground truth for these long-tail scenes, which classical SfM fails to provide. How do we bypass it? We break this bottleneck with two key contributions. First, we introduce MegaDepth-X, a large new dataset of scenes with high-quality 3D supervision. Second, we propose a new way to simulate difficult image sets for training. Project page: megadepth-x.github.io/
1
20
2
104
22,969
Noah Snavely retweeted
Yoav Artzi@yoavartziApr 24
We are organizing the @cornell_tech Frontiers of AI Summit to bring together industry and cutting-edge academic research. May 27 on the NYC Cornell Tech campus! Attendance info on the website: tech.cornell.edu/frontiers-o…
7
41
6,228
Noah Snavely retweeted
Gene Chou@gene_ch0uApr 22
Introducing CityRAG! We wanted video generative models to be grounded in the real world — if I’m in London, I want to look around and actually see Big Ben. CityRAG generates videos of cities featuring real buildings and roads, with arbitrary weather, people, and cars. 1/N page: cityrag.github.io paper: arxiv.org/abs/2604.19741
6
49
8
242
37,086
Noah Snavely retweeted
Neerja Thakkar@neerjathakkarApr 2
What’s the right representation for a world model? 3D, pixels, or something else? Excited to release our new paper “Forecasting Motion in the Wild” where we propose point tracks as tokens for generating complex non-rigid motion and behavior From @GoogleDeepmind @Berkeley_AI @TTIC_Connect
9
73
15
472
83,854
Noah Snavely retweeted
Qianqian Wang
@QianqianWang5Mar 10
Very excited to share our exploration of a new robotics foundation model at Rhoda AI. We train a causal video model from scratch, unlocking new capabilities for robust, long-horizon closed-loop robot control. Learn more: rhoda.ai/research/direct-vid…

Causal Video Models Are Data-Efficient Robot Policy Learners | Rhoda AI

At Rhoda AI, we are building towards generalist robotics. Our Direct Video-Action Model (DVA) reformulates robot policies as video generation, unlocking data-efficient task learning, scaling,...

rhoda.ai
Rhoda AI
@RhodaAIMar 10
To bring generalist intelligent robots to the real world, we have to overcome the data scarcity problem. At Rhoda, we are solving it by reformulating robot policies as video generation. Today, we introduce the Direct Video-Action Model (DVA)
8
1
107
10,542
In your post-ECCV haze, check out @Haian_Jin's really nice work on linear-time, feed-forward 3D reconstruction!
Haian Jin
@Haian_JinMar 5
Spatial reconstruction is a long-context problem: real scenes come with hundreds of images. But O(N²) transformer-based models don’t scale efficiently. Introducing: 🤐ZipMap (CVPR ’26): Linear-Time, Stateful 3D Reconstruction via Test-Time Training (TTT). ZipMap “zips” a large image collection into an implicit TTT scene state in a single linear-time operation. The state will then be decoded into spatial outputs, and can be queried efficiently for novel-view geometry and appearance (~100 FPS) ZipMap is not only much faster (>20× faster than VGGT), but also matches or surpasses the accuracy of all SOTA models.
2
4
43
6,678
Load more