Jerome Revaud (@JeromeRevaud) header
Jerome Revaud (@JeromeRevaud) profile picture

AI researcher :: Computer vision :: 3D vision :: Team Lead :: Project Lead :: Naver Labs Europe. Views are my own

Grenoble, France
Joined June 2023
Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)Photo by Jerome Revaud (@JeromeRevaud)
Not a paper, but certainly more important than a paper… at least to me
I have never felt so shaken. Children are dying of hunger. Falling one after another, like petals of a dying flower. Under our watch. How do you sleep, Presidents, Foreign Ministers, political leaders, diplomats, civil servants - doing nothing to stop Israel’s starvation campaign? PS: Talking politics while dodging genocide perfectly meets the definition of doing nothing.
1
2
24
1,602
Online reconstruction of ~400 sq meters in feed-forward regression only
Replying to @Vinc3nt_Leroy
RGB-SLAM becomes a single forward through the network with no optimization needed. We still need to only keep keyframes to avoid exploding the costs. We simply check the 3D overlap between views and memory, ensuring new keyframes discover enough novel content...
1
9
103
6,668
RT is a bit late but well
*3R posts are back! 🧵 Interested in SfM, RGB-SLAM or... both at the same time??? Come see MUSt3R @CVPR25 Friday morning, ExHall D Poster #82. Jerome and Boris will be there to present how we can adapt DUSt3R to multiple views via a memory mechanism (and maybe do parkour?!)
5
951
A teaser of the latest 3D DUSt3R based models we’re presenting at @cvprconference Discover MUSt3R & Pow3R, universal encoder DUNE + research in navigation, vizloc, segmentation & human motion understanding! All our #CVPR2025 papers are here ➡️ tinyurl.com/4z79ujce
2
33
1
229
14,279
Jerome Revaud retweeted
16
82
4,907
Happy to introduce our new CVPR paper—Pow3R: Empowering Unconstrained 3D Reconstruction with Scene and Camera Priors : arxiv.org/pdf/2503.17316 Inspired by the amazing DUSt3R, which reconstructs 3D from just two unposed images, we explored: (1/n)
3
40
2
209
17,141
Jerome Revaud retweeted
Pow3R: Empowering Unconstrained 3D Reconstruction with Camera and Scene Priors @wbjang11, @WeinzaepfelP, @Vinc3nt_Leroy, @LourdesAgapito, @JeromeRevaud tl;dr: DUSt3R with any optional input subset of camera intrinsics, pose and depthmaps arxiv.org/abs/2503.17316
9
42
2,296
Jerome Revaud retweeted
MUSt3R: Multi-view Network for Stereo 3D Reconstruction Yohann Cabon, Lucas Stoffl, Leonid Antsfeld, @kgcs96, Boris Chidlovskii, @JeromeRevaud, @Vinc3nt_Leroy tl;dr: make DUSt3R symmetric and iterative+multi-layer memory mechanism->multi-view DUSt3R arxiv.org/abs/2503.01661
1
10
108
5,027
Jerome Revaud retweeted
MASt3R-SLAM: it's easily the most robust dense monocular SLAM system I've ever seen, and also very accurate. Real-time, RGB only (no IMU), handles unknown focal length and zoom. More new videos coming soon @eric_dexheimer @rmurai0610. Paper and full info: edexheim.github.io/mast3r-sl…
Introducing MASt3R-SLAM, the first real-time monocular dense SLAM with MASt3R as a foundation. Easy to use like DUSt3R/MASt3R, from an uncalibrated RGB video it recovers accurate, globally consistent poses & a dense map. With @eric_dexheimer*, @AjdDavison (*Equal Contribution)
7
74
4
533
32,262
🚀 Our NeurIPS '24 work, Large Spatial Model (LSM), is here! LSM performs semantic 3D reconstruction in just 0.1s, processing unposed data via feed-forward 3D reconstruction. 👉It leverages large-scale 3D datasets with minimal annotations, defining a 3D latent space. We are continuously exploring how this explicit 3D representation can further enhance reasoning and robotic learning. 🔗 Try our online Gradio demo with your own data at largespatialmodel.github.io/ #NeurIPS2024 #3DReconstruction
3
64
6
306
44,076
Jerome Revaud retweeted
And colloquium is starting with @Vinc3nt_Leroy arguing we need one model to rule them all in 3D
This Post is from an account that no longer exists. Learn more
2
6
57
10,510
Not 3D vision but interesting work by Gabriela and colleagues about explainability in CV
We presented the poster at the eXCV workshop at #ECCV2024 . Thank you to people stopping by. If you missed it and you are interested in explainability in CV (#xAI) here is our poster.
4
989
Jerome Revaud retweeted
Excited to share MonST3R! -- a simple way to estimate geometry from unposed video of dynamic scene We achieve competitive results on several downstreams (video depth, camera pose) and believe this is a promising step toward feed-forward 4D reconstruction monst3r-project.github.io
22
138
33
719
131,868
Want to know more about single frame #mapfree localization benchmark winner🏆 #MASt3R & all our other presentations, posters & demos at #ECCV2024 this week? Here's a summary & schedule for the week➡️europe.naverlabs.com/blog/na…
8
16
2,454
Dr. Vincent Leroy from Naver Labs Europe presenting CroCo, Dust3r, Mast3r and how large foundational models could solve 3D tasks at Map-free Visual Relocalization workshop at #ECCV2024 .
1
4
52
3,039
I had to try. Not too bad despite the many non-Lambertian surfaces 😉
3
5
1
67
6,021
Trying to break MASt3R-SfM. Not quite 😁 Code at github.com/naver/mast3r/tree… , scales to 1000+ images on a small GPU (also code for MASt3R+GLOMAP, see demo_glomap.py)
1
10
1
73
18,846
MASt3R-SfM paper should be on arXiv tomorrow.
3
1
22
4,329