Alexander Winkler
146 posts
Alexander Winkler
@awinkler_
Research Scientist @Meta. RL, Robotics, Character Animation, Motion Tracking. Alumn @ETH
San Francisco, CA Katılım Ekim 2010
481 Takip Edilen1.7K Takipçiler
For any PhD students looking for a summer internship to work with me on RL, VLAs and humanoid robots in SF, please consider applying.
metacareers.com/profile/job_de…
English
Excited to share our latest work at Meta FAIR: introducing Meta Motivo (metamotivo.metademolab.com), a behavioral foundation model for controlling virtual physics-based humanoid agents.
English
Alexander Winkler retweetledi
Omnigrasp's poster will be at:
📍East Exhibit Hall A-C #4108
🕕 Fri 13 Dec 11 a.m. PST — 2 p.m. PST
at #NeurIPS2024
Due to visa issues, I couldn't be there😭. My friend @yus167 will be there to present. Feel free to reach out for a chat if you are interested in our work!
Zhengyi “Zen” Luo@zhengyiluo
Introducing Omnigrasp: Grasping Diverse Objects with Simulated Humanoids. With Omnigrasp, we show that we can control a humanoid equipped with dexterous hands to grasp diverse objects (>1200) and follow diverse trajectories, with one policy! 🌐: zhengyiluo.com/Omnigrasp/ 📜: zhengyiluo.com/Omnigrasp-Site…
English
English
Couldn't be more excited to share World Labs today, our journey towards Spatial Intelligence! We are making the world your oyster with @drfeifei, @BenMildenhall, @jcjohnss and a world class team - more to come!
World Labs@theworldlabs
Hello, world! We are World Labs, a spatial intelligence company building Large World Models (LWMs) to perceive, generate, and interact with the 3D world. Read more: worldlabs.ai/about
English
English
@zhengyiluo @RealityLabs It’s been a pleasure working with you. What a productive two years you’ve had, very impressed!💪
English
After two wonderful years visiting Meta @RealityLabs (via AI Mentorship Program) Codec Avatars Team, today is my last day!
What a journey! We started with the goal of leveraging physics simulation to aid egocentric pose estimation. We developed PHC, scaling humanoid motion imitation to tens of thousands of human motions; PULSE, a universal humanoid motion representation that can speed up RL (like a lot); SimXR, the task we set out to do: controlling simulated humanoids from a headset; Omnigrasp, grasping diverse objects with simulated humanoids. Hopefully, these works could inspire future work in character animation and robotics!
Thanks to my mentor @xuweipeng000, my advisor @kkitani, and my wonderful collaborators @awinkler_, Jing Huang, @jinkuncao, Josh Merel, @rawal_khirodkar.
English
Exciting work by Peizhuo and team👏Check it out.
Peizhuo Li@PeizhuoL
Want to #WalkTheDog in the metaverse? In our project at #SIGRAPH2024 with @blacksquirrel__ , Yuting Ye and @OlgaSorkineH, we develop an approach to learn a common 1D phase manifold from motion datasets across different morphologies, *without* any supervision (1/2)
English
Great work by Sebastian and team, check it out!
Sebastian Starke@blacksquirrel__
The code, dataset and demos for our #SIGGRAPH2024 work on generating embodied avatar movements with Codebook Matching are now available! Everything can be found on the project page, including a precompiled binary that you can play around with! github.com/sebastianstark…
English
Alexander Winkler retweetledi
Introducing Omnigrasp: Grasping Diverse Objects with Simulated Humanoids.
With Omnigrasp, we show that we can control a humanoid equipped with dexterous hands to grasp diverse objects (>1200) and follow diverse trajectories, with one policy!
🌐: zhengyiluo.com/Omnigrasp/
📜: zhengyiluo.com/Omnigrasp-Site…
English
Alexander Winkler retweetledi
🔥Excited to be in Seattle for #CVPR! I'll present #DiffAvatar, using diff. sim. to recover body & garment geometry + material parameters. 🚀Catch me at the poster session Wed, June 19, 10am, Arch 4A-E, #408! More info: people.csail.mit.edu/liyifei/diffav…。#MetaRealityLabs
English
Alexander Winkler retweetledi
Check our work at CVPR! A collaboration effort with @bharat_b7 @ Yuanlu Xu @awinkler_ @ Petr Kadlecek @SiyuTang3 @FedericaBogo
📍Oral 4C, Summit Flex Hall C
🗓️1.00-2.30pm, Thursday
📍Poster 182 @ poster session 4, Arch 4A-E
🗓️5.15-6.45pm, Thursday
English
Another impressive project led by Zen!💪
Zhengyi “Zen” Luo@zhengyiluo
#CVPR2024 Highlight🌟 SimXR: Real-Time Simulated Avatar from Head-Mounted Sensors Controlling simulated humanoids and estimate full-body pose using SLAM images and headset pose in real time. 🌐: zhengyiluo.github.io/SimXR 📜: arxiv.org/abs/2403.06862 🧑🏻💻: github.com/ZhengyiLuo/Sim… Input: four SLAM views from Quest 2 headset and 6DoF headset pose. Output: full-body humanoid control signals. Leveraging the laws of physics as a strong prior for egocentric pose estimation.
English
Alexander Winkler retweetledi
#ICLR2024 Spotlight🌟🌟🌟
PULSE: Physics-based Universal humanoid motion Latent SpacE/Representation
Code: github.com/ZhengyiLuo/PUL…
Site: zhengyiluo.com/PULSE
Paper: arxiv.org/abs/2310.04582
As a motion representation, PULSE is low dimensional (32), high coverage (99.8% of AMASS motion), can speed up downstream task w/ hierarchical RL, and can be randomly sampled as a generative model.
PULSE is akin to a foundation model for simulated humanoid control where downstream tasks ranging from simple locomotion, complex terrain traversal, to free-form motion tracking can all reuse this representation.
Zhengyi “Zen” Luo@zhengyiluo
Now that PHC's code is out... Introducing PULSE: Physics-based Universal motion Latent SpacE: 📜: arxiv.org/abs/2310.04582… 🌐: zhengyiluo.github.io/PULSE/ All downstream tasks here use the same pretrained latent space. (1/6)
English
Inspiring how far legged robotics has come. Congratulations to the team!
Robotic Systems Lab@leggedrobotics
🔥Exciting news 🤖 Our latest research by @HoellerDavid, @rdn_nikita, @2nisi in @SciRobotics unlocks new achievements: Unprecedented agility in quadrupedal robots, mastering locomotion, navigation, and perception through deep reinforcement learning! @NVIDIARobotics
English
Alexander Winkler retweetledi
Back in the lab this week updating Kinetic Diffusion, our AI powered dance performance, with new @StableDiffusion models. Speed and pixels have increased tremendously unlocking new opportunities. Here’s an experiment in “popping” new prompts in with connection to the movement ⬇️
English
Alexander Winkler retweetledi
While I am at it, also releasing the trained models for VR controllers tracking using PHC.
This task is a essentially a generalized version of motion imitation, where there are only 3 6DOF points to track (red dots) instead of 24.
Models are released at github.com/ZhengyiLuo/Per…
👇🏻Video here is a sequence from @awinkler_ 's QuestSim: arxiv.org/abs/2209.09391.
(This is basically bootleg QuestSim🤪)
English
Alexander Winkler retweetledi
2023 was a wild year for technological progress. At Meta, our two biggest long-term bets on the technologies of the future both make great leaps forward but more excitingly, they began to converge. I wrote about it here: meta.com/blog/quest/boz…
English
Impressive pace of progress📈. Congrats to the team!💪
Tesla Optimus@Tesla_Optimus
There’s a new bot in town 🤖 Check this out (until the very end)! tesla.com/ai
English