Cédric Rommel 🦋 (@ccrommel) header
Cédric Rommel 🦋 (@ccrommel) profile picture

Research Scientist at @Meta | AI and neural interfaces | Interested in data augmentation, generative models, geometric DL, brain decoding, human pose, …

Paris, France
Joined October 2015
Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)Photo by Cédric Rommel 🦋 (@ccrommel)
Cédric Rommel 🦋 retweeted
We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2. Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication. We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating. 🧵👇
657
2,039
1,376
13,977
6,023,318
Interested in ill-posed learning tasks, uncertainty prediction, conditional density estimation or multi-head deep neural networks ? In our new paper, accepted at #ICML24, we tackle these challenges by exploring the Winner-Takes-All (WTA) training scheme. [1/n]
1
10
2
35
5,592
Cédric Rommel 🦋 retweeted
📢Joblib release day📢 We just released joblib 1.4!! 🤠 To update it, just run: pip install -U joblib Some new features: return_generator support for dask backend, collecting the results as they complete, caching coroutines, ... 😎 Full release note: tinyurl.com/joblib-1-4
12
1
34
2,473
Cédric Rommel 🦋 retweeted
SALUDA: Surface-based Automotive Lidar Unsupervised Domain Adaptation. @Bjoern_Michele, @AlexandreBoulch, @gillespuy, Tuan-Hung Vu, Renaud Marlet, @nicolas_courty pdf: arxiv.org/pdf/2304.03251.pdf code: github.com/valeoai/SALUDA
1
6
18
814
Cédric Rommel 🦋 retweeted
For starters, transformers are equivariant to permutations. Like ConvNets are equivariant to translations. Equivariance to permutations is good when what you care about are relationships between input items at any distance.
There is a commonly held belief that Transformers have no inductive bias and that this bias is learned throughout the training process. This is not true. Transformers have very strong inductive biases.
17
59
2
496
113,575
Cédric Rommel 🦋 retweeted
For those of you who were wondering what I’ve been doing since I joined @Meta Reality Labs late 2022. Here is the first detailed scientific communication about our work. You can read the paper at: biorxiv.org/content/10.1101/…
1/7 For the past decade, our team at Meta Reality Labs (previously CTRL-labs) has been dedicated to developing a neuromotor interface. Our goal is to address the Human Computer Interaction challenge of providing effortless, intuitive, and efficient input to computers.
5
24
166
17,845
Cédric Rommel 🦋 retweeted
Are motion forecasting models ready for deployment? 🤔 In this new #ICRA2024 paper, we reveal several challenges that forecasting models face (and show they are not ready 🤫) 📜: arxiv.org/abs/2306.09281 💻: github.com/valeoai/MFEval 🧵🔽
1
9
2
22
3,187
Cédric Rommel 🦋 retweeted
One reason AI research has been progressing so fast is not just because of frequent and early publication of preprint on ArXiv and the exchange of open source code, but also because the ML/AI community has largely freed itself from the stranglehold of for-profit & paywalled scientific publishing.
Replying to @pmarca
Thankfully, ML/AI research is largely free of the commercial publishing stranglehold. Preprints are posted on ArXiv and OpenReview, short papers in top conferences like ICLR, NeurIPS, ICML, and a few others, and longer papers in JMLR and TMLR. All of these venues are open access and free for both readers and authors. ArXiv and OpenReview are supported by philanthropy (they need more), conferences by registration fees, and journals by nothing (it really costs essentially no money to run an online journal). The exceptions are the few folks who still think that publishing in Nature Machine Intelligence, Machine Learning Journal, and other for-profit journals is a good idea (looking at you, DeepMind 😠). Not surprisingly, many Nature MI papers are about hardware or applications of AI to the sciences, topics that are not well covered by "core" ML venues.
64
341
51
2,354
443,565
Cédric Rommel 🦋 retweeted
Let me tell a story about free books. In the mid 1990s, I started a project called DjVu at AT&T Labs. The purpose was to devise a new image compression format so that printed documents could be scanned at high resolution and distributed efficiently over the newly expanding Internet. The format was released in the late 90s/early 00s and adopted by websites li,e the Internet Archive. As a useful demonstration of the technology, I decided to scan and distribute the complete collection of proceedings of the Neural Information Processing conference (NIPS). I asked the publishers, Morgan Kaufman and MIT Press, for permission to do that. They agreed because they weren't making any revenue from past proceedings. We scanned the 13 volumes, OCRed and indexed all of the material, and put it up on a free website in 2000: nips.djvu.org/ This open-access repository turned out to be *extremely* useful to the machine learning research community. Around the same time, the ML community rebelled against commercial journal publishers and created JMLR, which was one of the first open-access and totally free journal. This also turned out to be enormously beneficial. Eventually, the NIPS conference stopped making printed proceedings and started hosting all the books on their website ( nips.cc ), including our scans. If you ever wondered why the ML/AI community embraced a culture of fast posting of preprint and open-access publications, that's it.
95
347
67
2,801
539,305
Cédric Rommel 🦋 retweeted
🍾POP-3D If you are interested in open-vocabulary segmentation in 3D using only images, come check our poster on Thursday morning! Poster #115 ⏲️Th. 14 Dec. 10:45 a.m. CST — 12:45 p.m. CST Thanks to @oriane_simeoni D.Hurych @SpyrosGidaris @abursuc @ptrkprz J.Sivic! 🥳
9
24
961
Cédric Rommel 🦋 retweeted
We are presenting our poster on links between GANs and Diffusion at Neurips. Don't be scared by the formalism, ⁦@jy_franceschi⁩ is awesome and will explain everything intuitively!
1
13
2
55
6,304
Cédric Rommel 🦋 retweeted
Evaluating the structure of cognitive tasks with transfer learning by @BAristimunha R. Camargo @Warvito @sylvcheva @agramfort @ccrommel tl;dr: build transferability maps between cognitive tasks for data efficient EEG decoding #neurips2023 @AI_for_Science cedricrommel.github.io/asset…
3
1
5
1,141
Cédric Rommel 🦋 retweeted
🍾POP-3D @NeurIPSConf If you want to discuss open-vocabulary 3D occupancy prediction from images only, reach out to @AVobecky who will present the poster on Thursday: Poster #115 @🗺️Great Hall & Hall B1+B2 (level 1) ⏲️Th. 14 Dec. 10:45 a.m. CST — 12:45 p.m. CST
POP-3D: Open-Vocabulary 3D Occupancy Prediction from Images by @AVobecky @oriane_simeoni D.Hurych @SpyrosGidaris @abursuc P.Pérez, J.Sivic tl;dr: open-vocab 3D semantic occupancy maps from multi-cam inputs w/ tri-modal self-supervised learning #neurips2023 vobecant.github.io/POP3D/
9
18
2,406
Cédric Rommel 🦋 retweeted
Super excited to be ✈️ @NeurIPConf next week to present our recent works on audio and multi-camera perception (supervised or self-supervised), diverse model soups and understanding of generative models. Find out more 👇 and come meet us at #neurips2023 valeoai.github.io/blog/2023/…
1
6
20
2,308
Cédric Rommel 🦋 retweeted
Should we create a motion for #CVPR #CVF so that quantitative comparisons to papers without publicly available source code cannot be asked by reviewers? (Note SIGGRAPH already has such a policy)
91%Yes (no code, no eval)
9%No
691 votes • Final results
15
9
4
52
32,722
Cédric Rommel 🦋 retweeted
🎉Crossing 10K citations 🎉 Thanks to all who've joined us on this quest! 🙏 ✅Follow us to get notified of our latest research: scholar.google.com/citations… We're proud of the team and collaborators, though we know citations are just a metric and not the end game
3
3
32
7,277
Cédric Rommel 🦋 retweeted
🚨💻 Join us at valeo.ai, Paris, for exciting research internships in computer vision & ML! We have several openings on diffusion models, scene understanding, motion forecasting, explainability, LLMs & more. If interested, contact me in PM valeoai.github.io/blog/
2
14
2
64
16,832
Cédric Rommel 🦋 retweeted
Really shocked today... My paper A*Net has been on arXiv for more than a year, and it was plagiarized in an ICLR submission. There is no reference to A*Net and more than half of the methodology is identical to my paper. @iclr_conf any solution? up: my A*Net. down: plagiarism.
24
66
4
601
163,688
Cédric Rommel 🦋 retweeted
Thanks all for coming to Paris @ICCVConference and making it such an amazing event. As warned, we brought the whole @valeoai family, interns and seniors alike, and it was a blast! We are so happy to share good papers, discussions and laughs. What an amazing community #ICCV2023
From interns to senior researchers, the @valeoai team will participate in mass at @ICCVConference and will be looking forward to welcoming you and talking about the exciting progress and ideas in the field. Check our #ICCV2023 works and come meet us 👇valeoai.github.io/blog/2023/…
7
29
6,209