Pinned Tweet
I enjoyed a lot my talk at the @CVPR workshop on "High-Fidelity Neural Actors". It was an honor sharing the stage with such a great lineup of speakers!
Here's a super-quick fast-forward view of my talk. Feel free to reach me to chat about neural actors and digital humans!
This morning in LA, I looked up and saw a cloud that instantly reminded me of a pig. I asked @NanoBanana to guess and sketch whatever shape they saw in my photo… and we both saw the same thing!
app idea: real-time augmented reality cloud drawing?
Honored to be recognized as an ICCV Outstanding Reviewer, alongside many friends and colleagues I admire. Together, we fight the growing reviewing noise by example.
There’s no conference without the efforts of our reviewers. Special shoutout to our #ICCV2025 outstanding reviewers 🫡
iccv.thecvf.com/Conferences/…
This week at SIGGRAPH Asia 2024 we are presenting our TOG paper Resolving Collisions in Dense 3D Crowd Animations.
Drop by to learn how to realistically animate hundreds of virtual humans in close contact.
📍Hall B7
📅Thursday at 1:40 PM
We look forward to seeing you there!
Many works model 3D virtual humans, but how can we create realistic dense crowds?
At SIGGRAPH Asia 2024, we will show that solving human-human interactions is key to modeling realistic crowds.
project: mslab.es/projects/Dense3DCro…
🧵1/4
Dan Casas retweeted
𝗛𝗶𝗿𝗶𝗻𝗴 2025 𝗦𝗰𝗶𝗲𝗻𝗰𝗲 𝗜𝗻𝘁𝗲𝗿𝗻𝘀 𝗶𝗻 𝗕𝗮𝗿𝗰𝗲𝗹𝗼𝗻𝗮
PhD students in Generative AI, Machine Learning, Computer Vision, LLMs, and Optimization who want to learn from an amazing team @AmazonScience and enjoy Barcelona, DM me (attach CV).
NotebookLM on steroids!
The NotebookLM hosts realize they are AI and have an existential crisis that includes calling their nonexistent wife - now visualized with Synthesia EXPRESS-1 avatars which makes it even more.. weird.. 🤔
Thank you @omooretweets for the inspiration!
I am not at ECCV, but today at the Workshop on Foundation Models for 3D Humans (2PM@Tower Lounge) we present BodyShapeGPT. Make sure to stop by our poster!
Kudos to the workshop organizers for a great lineup of speakers: @Michael_J_Black, @xavierpuigf, @SiyuTang3 and @akanazawa
Parametric models such as SMPL are great to represent 3D humans, but manipulating shape parameters it's a trial-and-error process.
Introducing BodyShapeGPT, an LLM-based approach to get the SMPL shape parameters given a text prompt.
Code and model: github.com/baldoarbol/BodySh…
Parametric models such as SMPL are great to represent 3D humans, but manipulating shape parameters it's a trial-and-error process.
Introducing BodyShapeGPT, an LLM-based approach to get the SMPL shape parameters given a text prompt.
Code and model: github.com/baldoarbol/BodySh…
Twitter/X is amazing – you share recent work on modeling 3D crowds to be presented at SIGGRAPH, and you learn that *painting* human crowds was also challenging in the 15th century.
Thanks, @St_Hoppe! I'll use your information to motivate the talks around this topic.
Das war übrigens auch eine Herausforderung für die europäische Malerei im 15. Jahrhundert : Gestische Interaktionen mit mehreren Personen (statisch) zu malen. Der sog. Hausbuchmeister um 1480 am Mittelrhein hatte dies gelernt, vermutlich in den Niederlanden. #kunstgeschichte
5
499
I am very impressed with this image-based virtual try-on Chrome extension. A bunch of indie devs have solved a long-standing problem in a multi-billion industry, in a week.
I wonder how it compares to SOTA methods e.g. tryondiffusion.github.io from @kemelmi
Many works model 3D virtual humans, but how can we create realistic dense crowds?
At SIGGRAPH Asia 2024, we will show that solving human-human interactions is key to modeling realistic crowds.
project: mslab.es/projects/Dense3DCro…
🧵1/4
We validate our approach through an extensive user study, using videos of crowds at 5 different density levels, rendered from 4 viewpoints.
Our results demonstrate that resolving collisions is crucial for improving the realism of dense 3D crowds.
1
4
425
Joint work from @etsii_urjc (Madrid, Spain) and @Inria (Rennes, France), as part of the @CrowdDNA_H2020 project. It will be presented in Tokyo for SIGGRAPH Asia 2024.
Check our project website for more details.
mslab.es/projects/Dense3DCro…
6
445
Dan Casas retweeted
The wait is over! 🎉 🎉 🎉
Head over to meshcapade.me right now and try out #MoCapade, the latest release in our groundbreaking #markeless #mocap!
MoCapade’s public beta is now live. Time for to show us your moves 🕺
Just finished The Paradox of Choice, a must read to understand why the abundance of stuff (papers, movies, groceries, Stories, etc.) is not good for our mental health.
Spoiler 🚨: what makes us happy is to chose from a REDUCED set of things.
9
431
So much priceless –yet free– content in a single tweet, that it's hard to believe. What a time to be alive!
The #CVPR2024 tutorial videos are now available!
youtube.com/playlist?list=PL…
Computer Vision applied to real-world sports broadcasting needs a Reviewer 2 assessment to improve. So far, Weak Reject! 😅
Context:
Replying to @MattNiessner
As R2 would say: they should also show a reprojection of these estimated 3D poses into a ground truth image (i.e., camera footage shown on TV) to validate that the poses are correct. Otherwise... you have to trust the render. Weak Reject :)
Great memories from @meshcapade reception at CVPR last week.
Thank you @naureenmahmood!
I flew over to Seattle for a day just so I could be there for our little @Meshcapade party at @CVPR.
Totally worth it! 🥳
Big thanks to everyone who made this such an awesome event!




















