gistvision/moca

Code and models of MOCA (Modular Object-Centric Approach) proposed in "Factorizing Perception and Policy for Interactive Instruction Following" (ICCV 2021). We address the task of long horizon instruction following with a modular architecture that decouples a task into visual perception and action policy prediction.

语言：C · GitHub 星标：40 · Fork：14

话题：instruction-following

在 GitHub 上查看原仓库