DreamX enables cross embodiment video action learning with compact codes
DreamX: Video Diffusion Model is a Universal Latent Action Learner
Computer Vision and Pattern RecognitionArtificial IntelligenceRobotics
Summary
Understanding actions in videos is hard because the same action can look very different depending on who or what is doing it. The authors introduce DreamX, a way to capture the essential parts of an action in a very small set of data, ignoring irrelevant visual details. This makes it easier to transfer knowledge about actions from humans to robots or between different environments. DreamX also helps in creating efficient policies for controlling actions based on these compact representations.
What this means in practice
- •For robotics engineers: Train robots to understand and replicate human actions by using DreamX’s shared compact action codes across embodiments.
- •For autonomous vehicle teams: Improve prediction and imitation of complex movements in varied environments by encoding actions in a compact, transferable latent space.
Authors
Jisoo Kim, Chaehyun Kim, Junsu Kim, Jiho Park, Seongchan Kim, Eunbeen Hong, Soowon Son, Fu-En Yang, Seungryong Kim
Abstract
Learning a universal action representation from video requires modeling dynamics that are consistent across different executions of the same action yet invariant to changes in embodiment and appearance. Existing latent action models infer actions from visual transitions, but their representations often entangle the underlying action with irrelevant visual variations, limiting transfer across embodiments. We introduce DreamX, which repurposes a pretrained video diffusion model as a universal latent action learner. Our approach extracts compact latent actions from the video model's rich spatiotemporal features, compressing each latent video frame by $\boldsymbol{300\times}$ ($\boldsymbol{1{,}200 \rightarrow 4}$ tokens). In addition, by leveraging the video model's generative prior, these latent actions capture action-relevant dynamics disentangled from visual appearance. The resulting latent space serves as a shared action interface across embodiments (human and robot), environments, and action parameterizations. On LARY-Bench, DreamX achieves state-of-the-art performance among latent action models. Moreover, actions encoded in one setting remain effective in another: the latent space supports any-to-any video transfer and zero-shot action transfer across diverse embodiments and environments. Furthermore, this shared interface also enables policy learning in a compact latent action space, reducing the computational cost of video-diffusion-based policies.