Sunmin Lee

🎉Excited to share our work #QuestEnvSim at #SIGGRAH2023 It learns to track environment-interaction motion from only Quest sensors and the surrounding geometry. - with Sebastian Starke, Yuting Ye, Jungdam Won, and Alexander Winkler (1/8)👇

🔥Phase 2 is heating up with a top submission from  @Jungnam_P and @JungdamW currently leading the 🦾 Manipulation track! Submission is open until Nov 6th!

🔥@Jungnam_P and @JungdamW hit the ball out of the park with their latest submission in the locomotion track of the #NeurIPS #MyoChallenge! Join the #MyoChallenge2023 to beat them! The challenge is still ongoing!

🎉 Super excited to share our work LAMA at #ICCV23! LAMA can synthesize a series of natural human motions including locomotion, scene interaction, and object manipulation in complex 3D scenes. w/ Hanbyul Joo @jhugestar A thread 🧵 (1/n) Project page: jiyewise.github.io/projects/LAMA/

Project page (8/8) For more, please check out Project: sunny-codes.github.io/projects/quest… Paper: arxiv.org/abs/2306.05666 Video: youtube.com/watch?v=HXkp3I… Hope to see you at #SIGGRAPH2023!😉

Limitations (7/8) - Separate policies are trained for each interaction type. A single tracker covering broader motion repertoires would be ideal. - Policy may fail to recover from tracking drift. The simulated character can fall over like in other physics-based systems.

🎉Excited to share our work #QuestEnvSim at #SIGGRAH2023 It learns to track environment-interaction motion from only Quest sensors and the surrounding geometry. - with Sebastian Starke, Yuting Ye, Jungdam Won, and Alexander Winkler (1/8)👇

thread test

Hello, world! 🌏