NEVER LOOK BACK! Introducing LEDGER: persistent 3D object memory from egocentric videos. Remember where things were, when they moved, and what they held -> then answer questions without looking back at the video!

This was a fun project led by @Shra_1_C along with William Paul, and Johns Hopkins colleagues @suchisaria and Rama Chellappa Thread with details from @Shra_1_C below

object permanence for agents is so underrated. does ledger track things that leave the frame for hours, like keys that got moved by someone else while the wearer was out?

Persistent 3D memory could be a huge step for egocentric AI systems

Persistent 3D memory could be a major leap for embodied AI reasoning

Actually solid, also real-time implementation will make this🤌🏼

Smart approach to remembering object history

Persistent 3D object memory from egocentric video is wild answering where things were without rewatching is huge

Egocentric video memory that actually persists