Object-Centric Latent Action Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Klepach, Albina, Nikulin, Alexander, Zisman, Ilya, Tarasov, Denis, Derevyagin, Alexander, Polubarov, Andrei, Lyubaykin, Nikita, Kiselev, Igor, Kurenkov, Vladislav
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915740994502656
author Klepach, Albina
Nikulin, Alexander
Zisman, Ilya
Tarasov, Denis
Derevyagin, Alexander
Polubarov, Andrei
Lyubaykin, Nikita
Kiselev, Igor
Kurenkov, Vladislav
author_facet Klepach, Albina
Nikulin, Alexander
Zisman, Ilya
Tarasov, Denis
Derevyagin, Alexander
Polubarov, Andrei
Lyubaykin, Nikita
Kiselev, Igor
Kurenkov, Vladislav
contents Leveraging vast amounts of unlabeled internet video data for embodied AI is currently bottlenecked by the lack of action labels and the presence of action-correlated visual distractors. Although recent latent action policy optimization (LAPO) has shown promise in inferring proxy action labels from visual observations, its performance degrades significantly when distractors are present. To address this limitation, we propose a novel object-centric latent action learning framework that centers on objects rather than pixels. We leverage self-supervised object-centric pretraining to disentangle the movement of the agent and distracting background dynamics. This allows LAPO to focus on task-relevant interactions, resulting in more robust proxy-action labels, enabling better imitation learning and efficient adaptation of the agent with just a few action-labeled trajectories. We evaluated our method in eight visually complex tasks across the Distracting Control Suite (DCS) and Distracting MetaWorld (DMW). Our results show that object-centric pretraining mitigates the negative effects of distractors by 50%, as measured by downstream task performance: average return (DCS) and success rate (DMW).
format Preprint
id arxiv_https___arxiv_org_abs_2502_09680
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Object-Centric Latent Action Learning
Klepach, Albina
Nikulin, Alexander
Zisman, Ilya
Tarasov, Denis
Derevyagin, Alexander
Polubarov, Andrei
Lyubaykin, Nikita
Kiselev, Igor
Kurenkov, Vladislav
Computer Vision and Pattern Recognition
Artificial Intelligence
Leveraging vast amounts of unlabeled internet video data for embodied AI is currently bottlenecked by the lack of action labels and the presence of action-correlated visual distractors. Although recent latent action policy optimization (LAPO) has shown promise in inferring proxy action labels from visual observations, its performance degrades significantly when distractors are present. To address this limitation, we propose a novel object-centric latent action learning framework that centers on objects rather than pixels. We leverage self-supervised object-centric pretraining to disentangle the movement of the agent and distracting background dynamics. This allows LAPO to focus on task-relevant interactions, resulting in more robust proxy-action labels, enabling better imitation learning and efficient adaptation of the agent with just a few action-labeled trajectories. We evaluated our method in eight visually complex tasks across the Distracting Control Suite (DCS) and Distracting MetaWorld (DMW). Our results show that object-centric pretraining mitigates the negative effects of distractors by 50%, as measured by downstream task performance: average return (DCS) and success rate (DMW).
title Object-Centric Latent Action Learning
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2502.09680