Meta’s OK-Robot performs zero-shot pick-and-drop in unseen environments

OK-Robot combines three primary subsystems: an open-vocabulary object navigation module, an RGB-D grasping module and a dropping heuristic system.

ai ai newsletter automation business engineering & technology environments innovator spotlight newsletter meta natural language processing natural language querying navigation new york university programming & development rgb-d robot robotics science vision language model vision transformer model vit vlms zero-shot