Disentangled Object-Centric Image Representation for Robotic Manipulation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915198587109376 |
|---|---|
| author | Emukpere, David Deffayet, Romain Wu, Bingbing Brégier, Romain Niemaz, Michael Meunier, Jean-Luc Proux, Denys Renders, Jean-Michel Kim, Seungsu |
| author_facet | Emukpere, David Deffayet, Romain Wu, Bingbing Brégier, Romain Niemaz, Michael Meunier, Jean-Luc Proux, Denys Renders, Jean-Michel Kim, Seungsu |
| contents | Learning robotic manipulation skills from vision is a promising approach for developing robotics applications that can generalize broadly to real-world scenarios. As such, many approaches to enable this vision have been explored with fruitful results. Particularly, object-centric representation methods have been shown to provide better inductive biases for skill learning, leading to improved performance and generalization. Nonetheless, we show that object-centric methods can struggle to learn simple manipulation skills in multi-object environments. Thus, we propose DOCIR, an object-centric framework that introduces a disentangled representation for objects of interest, obstacles, and robot embodiment. We show that this approach leads to state-of-the-art performance for learning pick and place skills from visual inputs in multi-object environments and generalizes at test time to changing objects of interest and distractors in the scene. Furthermore, we show its efficacy both in simulation and zero-shot transfer to the real world. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_11565 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Disentangled Object-Centric Image Representation for Robotic Manipulation Emukpere, David Deffayet, Romain Wu, Bingbing Brégier, Romain Niemaz, Michael Meunier, Jean-Luc Proux, Denys Renders, Jean-Michel Kim, Seungsu Computer Vision and Pattern Recognition Robotics Learning robotic manipulation skills from vision is a promising approach for developing robotics applications that can generalize broadly to real-world scenarios. As such, many approaches to enable this vision have been explored with fruitful results. Particularly, object-centric representation methods have been shown to provide better inductive biases for skill learning, leading to improved performance and generalization. Nonetheless, we show that object-centric methods can struggle to learn simple manipulation skills in multi-object environments. Thus, we propose DOCIR, an object-centric framework that introduces a disentangled representation for objects of interest, obstacles, and robot embodiment. We show that this approach leads to state-of-the-art performance for learning pick and place skills from visual inputs in multi-object environments and generalizes at test time to changing objects of interest and distractors in the scene. Furthermore, we show its efficacy both in simulation and zero-shot transfer to the real world. |
| title | Disentangled Object-Centric Image Representation for Robotic Manipulation |
| topic | Computer Vision and Pattern Recognition Robotics |
| url | https://arxiv.org/abs/2503.11565 |