Opinions are my own. AI Research Scientist : PhD, Computer Vision @ Sorbonne University & Valeo.ai

Joined October 2023
13 Photos and videos
Need pixel-level features from your backbone (DINOv3, CLIP, RADIO, FRANCA...)? 🚀Introducing NAF: A universal, zero-shot feature upsampler. It turns low-res ViT features into pixel-perfect maps. -⚡ Model-agnostic -🥇 SoTA results -🚀 4× faster than SoTA -📈 Scales up to 2K res.
4
21
95
19,299
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
"What will happen to Europe if it keeps ignoring AI?" Three American labs each (!!) operate more AI compute than all of Europe combined. Today we're launching Europe 2031: a story of what might happen if that doesn't change.
29
101
419
205,310
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
🇫🇷🇺🇦 Tellement hâte d’être le 14 juillet et voir la tête certains 🙃 L’aviation Ukrainienne sera présente et un Mirage 2000 aux couleurs de l’armée de l’air Ukrainienne volera aux côtés de leurs homologues (et formateurs) 🇫🇷 Belle décision !
171
495
2,618
28,622
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
Come see our poster tomorrow afternoon at #ICRA2026 to talk about 3D representation alignment of diffusion models for LiDAR generation. 📍196, Hall C 🗓️ Wednesday, 3:00-4:30 PM
📢 New preprint! We introduce R3DPA - a LiDAR scene generator that: • Transfers RGB-pretrained generative priors to 3D • Aligns with self-supervised LiDAR features • Enables object inpainting & scene mixing at inference • Sets SOTA on KITTI-360
1
5
7
849
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
1/n New paper - V-GIFT 🎁 Self-supervised tasks like rotation prediction or colorization were big in 2018. Do they still matter? Yes. We turn them into visual instruction tuning data for MLLMs. Result: models rely more on the image and perform better on vision tasks 👀
3
23
86
11,955
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
Our visiting PhD student @bliberatori_ gave us a great talk on Video understanding beyond trimmed clips: temporal and comparative reasoning. Make sure to have a look at her excellent works in this area: benedettaliberatori.github.i…
1
6
25
1,620
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
Thrilled to announce that R3DPA has been accepted at ICRA 2026 in Vienna. Can’t wait to share our work during June 1–5! #ICRA2026
📢 New preprint! We introduce R3DPA - a LiDAR scene generator that: • Transfers RGB-pretrained generative priors to 3D • Aligns with self-supervised LiDAR features • Enables object inpainting & scene mixing at inference • Sets SOTA on KITTI-360
1
6
11
1,294
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
📢 New preprint! We introduce R3DPA - a LiDAR scene generator that: • Transfers RGB-pretrained generative priors to 3D • Aligns with self-supervised LiDAR features • Enables object inpainting & scene mixing at inference • Sets SOTA on KITTI-360
2
35
177
11,721
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
1/🧵 Q: Can we have both a simple and SOTA architecture in autonomous driving? R: Yes! 😍 Introducing Driving on Registers (DrivoR): a pure Transformer backbone that achieves SOTA results in NAVSIM v1 / v2 and closed-loop HUGSIM evaluation. Here is how 👇
1
13
29
10,113
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
Oh yes let’s fly 30,000 people to Sydney! Crazy what irresponsible locations they choose for “AI conferences”, which in reality are just glorified holidays
#NeurIPS2026 will be held in Sydney, Australia! #ICML2017 was also in Sydney and was an absolute blast
33
4
379
182,978
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
By equating the European Union with the Third Reich—a regime responsible for systematic genocide, concentration camps, and World War II—Elon Musk not only trivializes the unspeakable horrors of National Socialism but also devalues ​​any serious debate about regulation and democracy. It's like comparing a speed limit to the Gestapo: absurd, tasteless, and intellectually bankrupt. The EU is not a totalitarian monstrosity but an alliance of 27 democracies that emerged after the horrors of the 20th century—including the Holocaust unleashed by Germany—to ensure peace, prosperity, and human rights. Billions of euros are invested in climate protection, social welfare systems, and civil liberties, while Musk's X (formerly Twitter) has degenerated into a digital madhouse under his leadership: hate speech flourishes, fake news runs rampant, and journalists are silenced. Who is the censor here? Musk, who portrays himself as a freedom fighter while taking legal action in Europe for ignoring EU laws like the Digital Services Act—laws designed to combat the very kind of propaganda he now promotes—is making a comparison that is not only historically inaccurate but cynical. It ignores the fact that the EU was built from the ruins of the Third Reich to make such dictatorships impossible forever. Musk, a billionaire who benefits from US subsidies and grew up in apartheid-era South Africa, should focus on his own hubris. Instead of resorting to cheap provocations to grab attention, he could build a spaceship that doesn't explode—or at least a tweet that doesn't cross the line of good taste. Such pronouncements don't make him a hero; they make him a populist jester. It's time he stopped instrumentalizing history and learned what respect for victims truly means.
2,455
2,690
12,837
443,093
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
3 Dec 2025
🚀Thrilled to introduce JAFAR—a lightweight, flexible, plug-and-play module that upsamples features from any Foundation Vision Encoder to any desired output resolution (1/n) Paper : arxiv.org/abs/2506.11136 Project Page: jafar-upsampler.github.io Github: github.com/PaulCouairon/JAFA…
1
1
5
128
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
Come see JAFAR do some magic tricks today December 3rd at 11:00 poster #5405! This is the coolest thing you'll see today!
🚀Thrilled to introduce JAFAR—a lightweight, flexible, plug-and-play module that upsamples features from any Foundation Vision Encoder to any desired output resolution (1/n) Paper : arxiv.org/abs/2506.11136 Project Page: jafar-upsampler.github.io Github: github.com/PaulCouairon/JAFA…
1
3
13
545
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
If you are interested in 3D or LiDAR generation, come check out our poster (with @Mickael_Chen @nemka_ and Renaud Marlet) this afternoon (ses. 2) LOGen: Toward Lidar Object Generation by Point Diffusion arxiv.org/abs/2412.07385 fully open sourced here: github.com/valeoai/LOGen

4
12
1,044
Loïck Chambon (PhD in CV) - 🇺🇦🇮🇷🇪🇺 retweeted
Humanity is at a turning point. I am launching UMA to build general-purpose mobile and humanoid robots from Europe. Proud to start with people I admired for years, and grateful for all your support! Reach out to us @UMA_Robots ❤️
168
151
1,321
447,635
🎉 Try NAF live — now on Hugging Face! 🤗 🖼️ Upload any image 🤖 Select any VFM (DINOv2, CLIP, RADIO, FRANCA...) 📐 Choose any resolution ⚡ Instant upsampling Try it now: huggingface.co/spaces/LChamb…
1
5
14
767
🚀 Real-Time Performance Because NAF is: 1️⃣ Model-agnostic 2️⃣ Ultra-lightweight (600k params) ...the demo runs incredibly fast! ⚡ And NAF is still 3️⃣ SoTA ! Experience how we turn low-res feature maps into high-fidelity outputs on the fly.
1
1
84
👀 Can’t wait to see what you all see! Tag me if you make interesting visualizations or discover surprising patterns. Paper: huggingface.co/papers/2511.1… Code: github.com/valeoai/NAF Original thread: x.com/LoickCh/status/1993263…
Need pixel-level features from your backbone (DINOv3, CLIP, RADIO, FRANCA...)? 🚀Introducing NAF: A universal, zero-shot feature upsampler. It turns low-res ViT features into pixel-perfect maps. -⚡ Model-agnostic -🥇 SoTA results -🚀 4× faster than SoTA -📈 Scales up to 2K res.
1
50
Need pixel-level features from your backbone (DINOv3, CLIP, RADIO, FRANCA...)? 🚀Introducing NAF: A universal, zero-shot feature upsampler. It turns low-res ViT features into pixel-perfect maps. -⚡ Model-agnostic -🥇 SoTA results -🚀 4× faster than SoTA -📈 Scales up to 2K res.
4
21
95
19,299