David Picard
@davidpicard.bsky.social
Professor of Computer Vision/Machine Learning at Imagine/LIGM, École nationale des Ponts et Chaussées @ecoledesponts.bsky.social Music & overall happiness 🌳🪻 Born well below 350ppm 😬 mostly silly personal views
📍Paris 🔗 https://davidpicard.github.io/
📍Paris 🔗 https://davidpicard.github.io/
Pinned
It's the first of the year. There is beauty in this world, it's not even hidden, let's focus on making it grow.
I'll make a gallery when I have time (so, never?)
February 9, 2026 at 9:50 AM
I'll make a gallery when I have time (so, never?)
Reposted by David Picard
💬 AI for Good Webinar Series 'From Molecules to Models' - today with Tomasz Trzcinski (IDEAS Research Institute 🇵🇱) on 'Zero-waste machine learning'.
📅 Mon, Feb 9th
🕓 16:00-17:00 CET
🔗 Registration: https://bit.ly/4rxevLX
ℹ️ More info: https://bit.ly/47Qclim
📅 Mon, Feb 9th
🕓 16:00-17:00 CET
🔗 Registration: https://bit.ly/4rxevLX
ℹ️ More info: https://bit.ly/47Qclim
February 9, 2026 at 7:45 AM
💬 AI for Good Webinar Series 'From Molecules to Models' - today with Tomasz Trzcinski (IDEAS Research Institute 🇵🇱) on 'Zero-waste machine learning'.
📅 Mon, Feb 9th
🕓 16:00-17:00 CET
🔗 Registration: https://bit.ly/4rxevLX
ℹ️ More info: https://bit.ly/47Qclim
📅 Mon, Feb 9th
🕓 16:00-17:00 CET
🔗 Registration: https://bit.ly/4rxevLX
ℹ️ More info: https://bit.ly/47Qclim
Not entirely satisfied...
February 8, 2026 at 9:39 PM
Not entirely satisfied...
Reposted by David Picard
#KostasThoughts: Stay organized and keep intermediate, raw project results. You never know when you, or your supervisor 😉, will need them later for a talk.
February 8, 2026 at 7:00 PM
#KostasThoughts: Stay organized and keep intermediate, raw project results. You never know when you, or your supervisor 😉, will need them later for a talk.
Exploring more weirdness.
February 8, 2026 at 10:10 AM
Exploring more weirdness.
Reposted by David Picard
OH "the S in MCP stands for security"
February 8, 2026 at 9:34 AM
OH "the S in MCP stands for security"
Reposted by David Picard
✨ New 3D pose estimation method from my lab! #FMPose3D allows for monocular (i.e. single camera) 2D➡️3D 🔥
Led by Ti Wang & w/ Xiaohang Yu #FMPose3D is SOTA on human & animal 3D benchmarks, & will be integrated into @deeplabcut.bsky.social ⬇️
📝 arxiv.org/abs/2602.05755
➡️ xiu-cs.github.io/FMPose3D/
Led by Ti Wang & w/ Xiaohang Yu #FMPose3D is SOTA on human & animal 3D benchmarks, & will be integrated into @deeplabcut.bsky.social ⬇️
📝 arxiv.org/abs/2602.05755
➡️ xiu-cs.github.io/FMPose3D/
February 8, 2026 at 7:08 AM
✨ New 3D pose estimation method from my lab! #FMPose3D allows for monocular (i.e. single camera) 2D➡️3D 🔥
Led by Ti Wang & w/ Xiaohang Yu #FMPose3D is SOTA on human & animal 3D benchmarks, & will be integrated into @deeplabcut.bsky.social ⬇️
📝 arxiv.org/abs/2602.05755
➡️ xiu-cs.github.io/FMPose3D/
Led by Ti Wang & w/ Xiaohang Yu #FMPose3D is SOTA on human & animal 3D benchmarks, & will be integrated into @deeplabcut.bsky.social ⬇️
📝 arxiv.org/abs/2602.05755
➡️ xiu-cs.github.io/FMPose3D/
Interesting video showing the gamers' viewpoint on recent video generative models: www.youtube.com/watch?v=Xsae...
I think this is not going to age well. Here is a prediction I feel confident about: in 10 years, most big title video game will rely on generative world models.
I think this is not going to age well. Here is a prediction I feel confident about: in 10 years, most big title video game will rely on generative world models.
Is The Gaming Industry COOKED?
YouTube video by gameranx
www.youtube.com
February 7, 2026 at 7:26 PM
Interesting video showing the gamers' viewpoint on recent video generative models: www.youtube.com/watch?v=Xsae...
I think this is not going to age well. Here is a prediction I feel confident about: in 10 years, most big title video game will rely on generative world models.
I think this is not going to age well. Here is a prediction I feel confident about: in 10 years, most big title video game will rely on generative world models.
Reposted by David Picard
February 7, 2026 at 9:01 AM
Reposted by David Picard
I asked Claude Code to build Claude Code. You won’t believe what happened next.
Secrets of Intelligence Services
I asked Claude code to build Claude code. You won’t believe what happened next.
www.argmin.net
February 6, 2026 at 3:26 PM
I asked Claude Code to build Claude Code. You won’t believe what happened next.
Reposted by David Picard
We are recruiting four positions connected to Machine Learning, Statistical Learning, and AI for Science in the Applied Mathematics department at École polytechnique. Join our vibrant community at IP Paris and Hi! Paris IA center. List below🧵 tinyurl.com/3jpw9t26
Calliopé
tinyurl.com
February 6, 2026 at 7:56 AM
We are recruiting four positions connected to Machine Learning, Statistical Learning, and AI for Science in the Applied Mathematics department at École polytechnique. Join our vibrant community at IP Paris and Hi! Paris IA center. List below🧵 tinyurl.com/3jpw9t26
New day, new experiments, new images
(with uin8 overflow artifacts, will be fixed for the next iteration)
(with uin8 overflow artifacts, will be fixed for the next iteration)
February 5, 2026 at 9:52 PM
New day, new experiments, new images
(with uin8 overflow artifacts, will be fixed for the next iteration)
(with uin8 overflow artifacts, will be fixed for the next iteration)
Reposted by David Picard
For absolutely no reason, let me remind people of this banger of a paper by @caroartc.bsky.social
doi.org/10.1016/j.jp...
doi.org/10.1016/j.jp...
January 30, 2026 at 10:22 AM
For absolutely no reason, let me remind people of this banger of a paper by @caroartc.bsky.social
doi.org/10.1016/j.jp...
doi.org/10.1016/j.jp...
That's a cool place to be!
There is an Associate Professor position in CS at ENS Lyon, with potential integration in my team, starting in sept 2026: DM me in interested!
Details at www.ens-lyon.fr/LIP/images/P...
Details at www.ens-lyon.fr/LIP/images/P...
www.ens-lyon.fr
February 5, 2026 at 2:02 PM
That's a cool place to be!
Reposted by David Picard
There is an Associate Professor position in CS at ENS Lyon, with potential integration in my team, starting in sept 2026: DM me in interested!
Details at www.ens-lyon.fr/LIP/images/P...
Details at www.ens-lyon.fr/LIP/images/P...
www.ens-lyon.fr
February 5, 2026 at 9:04 AM
There is an Associate Professor position in CS at ENS Lyon, with potential integration in my team, starting in sept 2026: DM me in interested!
Details at www.ens-lyon.fr/LIP/images/P...
Details at www.ens-lyon.fr/LIP/images/P...
Reposted by David Picard
Oh, that is definitely deep learning. I would even argue that deep learning started like this. These are my lecture slides for the "history of deeplearning":
1/3
1/3
February 5, 2026 at 11:02 AM
Oh, that is definitely deep learning. I would even argue that deep learning started like this. These are my lecture slides for the "history of deeplearning":
1/3
1/3
Reposted by David Picard
Ceci est censé être un document d'initiation pour un large public qui pourrait sinon être effrayé.
complexityzoo.net/Petting_Zoo
complexityzoo.net/Petting_Zoo
complexityzoo.net
February 5, 2026 at 8:28 AM
Ceci est censé être un document d'initiation pour un large public qui pourrait sinon être effrayé.
complexityzoo.net/Petting_Zoo
complexityzoo.net/Petting_Zoo
Let me continue the"or not" with this recent paper: arxiv.org/abs/2602.02493
Simple but very effective idea building on top of JiT: because you're predicting x directly, you can add perceptual losses on top of flow matching. In the paper, they use a "DINO perceptual loss", and I'm going to argue...
Simple but very effective idea building on top of JiT: because you're predicting x directly, you can add perceptual losses on top of flow matching. In the paper, they use a "DINO perceptual loss", and I'm going to argue...
February 5, 2026 at 8:45 AM
Let me continue the"or not" with this recent paper: arxiv.org/abs/2602.02493
Simple but very effective idea building on top of JiT: because you're predicting x directly, you can add perceptual losses on top of flow matching. In the paper, they use a "DINO perceptual loss", and I'm going to argue...
Simple but very effective idea building on top of JiT: because you're predicting x directly, you can add perceptual losses on top of flow matching. In the paper, they use a "DINO perceptual loss", and I'm going to argue...
Reposted by David Picard
Bibliography cleanup is known to be harder than AGI. Same with successfully connecting a laptop with a projector. It's what we humans will be doing long after the robot takeover. 😇
February 4, 2026 at 10:56 PM
Bibliography cleanup is known to be harder than AGI. Same with successfully connecting a laptop with a projector. It's what we humans will be doing long after the robot takeover. 😇
Reposted by David Picard
Zhao et al., "Sparsely Supervised Diffusion"
Simple, but seemingly effective idea. Just randomly masking your diffusion supervision seems to lead to less overfitting (of course?). Not to be confused with masked diffusion, this is simply during training.
Simple, but seemingly effective idea. Just randomly masking your diffusion supervision seems to lead to less overfitting (of course?). Not to be confused with masked diffusion, this is simply during training.
February 4, 2026 at 8:57 PM
Zhao et al., "Sparsely Supervised Diffusion"
Simple, but seemingly effective idea. Just randomly masking your diffusion supervision seems to lead to less overfitting (of course?). Not to be confused with masked diffusion, this is simply during training.
Simple, but seemingly effective idea. Just randomly masking your diffusion supervision seems to lead to less overfitting (of course?). Not to be confused with masked diffusion, this is simply during training.
Reposted by David Picard
Anthropic’s Super Bowl ad which criticizes AI chatbots that run ads (aka ChatGPT) just dropped. They aren’t pulling any punches and I love the song choice.
February 4, 2026 at 8:14 PM
Anthropic’s Super Bowl ad which criticizes AI chatbots that run ads (aka ChatGPT) just dropped. They aren’t pulling any punches and I love the song choice.
