ImagiNav: Scalable Embodied Navigation via Generative Visual Prediction and Inverse Dynamics
Paper • 2603.13833 • Published
ImagiNav
ImagiNav is a scalable embodied navigation framework that synthesizes future egocentric videos conditioned on language instructions and uses an inverse dynamics model to extract metric trajectories for execution. This repository hosts the model described in the paper ImagiNav: Scalable Embodied Navigation via Generative Visual Prediction and Inverse Dynamics.
- Project page: https://j1dan.github.io/ImagiNav
- Code: https://github.com/J1dan/ImagiNav-release