Object-Centric Representations Improve Policy Generalization in Robot Manipulation
Fuente:
arXiv
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915291447951360 |
|---|---|
| author | Chapin, Alexandre Machado, Bruno Dellandrea, Emmanuel Chen, Liming |
| author_facet | Chapin, Alexandre Machado, Bruno Dellandrea, Emmanuel Chen, Liming |
| contents | Visual representations are central to the learning and generalization capabilities of robotic manipulation policies. While existing methods rely on global or dense features, such representations often entangle task-relevant and irrelevant scene information, limiting robustness under distribution shifts. In this work, we investigate object-centric representations (OCR) as a structured alternative that segments visual input into a finished set of entities, introducing inductive biases that align more naturally with manipulation tasks. We benchmark a range of visual encoders-object-centric, global and dense methods-across a suite of simulated and real-world manipulation tasks ranging from simple to complex, and evaluate their generalization under diverse visual conditions including changes in lighting, texture, and the presence of distractors. Our findings reveal that OCR-based policies outperform dense and global representations in generalization settings, even without task-specific pretraining. These insights suggest that OCR is a promising direction for designing visual systems that generalize effectively in dynamic, real-world robotic environments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_11563 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Object-Centric Representations Improve Policy Generalization in Robot Manipulation Chapin, Alexandre Machado, Bruno Dellandrea, Emmanuel Chen, Liming Robotics Artificial Intelligence Image and Video Processing Visual representations are central to the learning and generalization capabilities of robotic manipulation policies. While existing methods rely on global or dense features, such representations often entangle task-relevant and irrelevant scene information, limiting robustness under distribution shifts. In this work, we investigate object-centric representations (OCR) as a structured alternative that segments visual input into a finished set of entities, introducing inductive biases that align more naturally with manipulation tasks. We benchmark a range of visual encoders-object-centric, global and dense methods-across a suite of simulated and real-world manipulation tasks ranging from simple to complex, and evaluate their generalization under diverse visual conditions including changes in lighting, texture, and the presence of distractors. Our findings reveal that OCR-based policies outperform dense and global representations in generalization settings, even without task-specific pretraining. These insights suggest that OCR is a promising direction for designing visual systems that generalize effectively in dynamic, real-world robotic environments. |
| title | Object-Centric Representations Improve Policy Generalization in Robot Manipulation |
| topic | Robotics Artificial Intelligence Image and Video Processing |
| url | https://arxiv.org/abs/2505.11563 |