Fei-Fei Li (@drfeifei)

2024-10-11 | โค๏ธ 294 | ๐Ÿ” 39


How do you effectively scale up real to sim data for robotic learning? We propose a new idea called โ€œdigital cousinsโ€ to simultaneously reduce the cost of real to sim generation while improving generalizability in learning ๐Ÿฆพ๐Ÿ˜


์ธ์šฉ ํŠธ์œ—

Tianyuan Dai (@RogerDai1217)

Why hand-engineer digital twins when digital cousins are free?

Check out ACDC: Automated Creation of Digital Cousins ๐Ÿ‘ญ for Robust Policy Learning, accepted at @corl2024! ๐ŸŽ‰

๐Ÿ“ธ Single image โ†’ ๐Ÿก Interactive scene

โฉ Fully automatic (no annotations needed!) ๐Ÿฆพ Robot policies deployed zero-shot in original scene

๐ŸŒ: https://t.co/GRsPGLuEaV

์›๋ณธ ํŠธ์œ—

๐ŸŽฌ ์˜์ƒ

Tags

domain-robotics domain-genai