Behavior Cloning for Active Perception with Low-Resolution Egocentric Vision
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916011847974912 |
|---|---|
| author | Bilic, Anthony Chen, Chen Bölöni, Ladislau |
| author_facet | Bilic, Anthony Chen, Chen Bölöni, Ladislau |
| contents | We investigate whether behavior cloning is sufficient to produce active perception in a structured object-finding task. A low-cost robot arm equipped with a wrist-mounted egocentric RGB camera must reposition to center a partially visible plant before triggering a grasp signal, requiring actions that improve future observations. The model predicts joint commands directly from low-resolution RGB images under closed-loop control. We show that low-resolution egocentric vision is sufficient for reliable task completion and that predicting relative joint deltas substantially outperforms absolute joint position prediction in our setting. These results demonstrate that visually grounded active perception can emerge from behavior cloning in a reproducible setting. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_14106 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Behavior Cloning for Active Perception with Low-Resolution Egocentric Vision Bilic, Anthony Chen, Chen Bölöni, Ladislau Robotics We investigate whether behavior cloning is sufficient to produce active perception in a structured object-finding task. A low-cost robot arm equipped with a wrist-mounted egocentric RGB camera must reposition to center a partially visible plant before triggering a grasp signal, requiring actions that improve future observations. The model predicts joint commands directly from low-resolution RGB images under closed-loop control. We show that low-resolution egocentric vision is sufficient for reliable task completion and that predicting relative joint deltas substantially outperforms absolute joint position prediction in our setting. These results demonstrate that visually grounded active perception can emerge from behavior cloning in a reproducible setting. |
| title | Behavior Cloning for Active Perception with Low-Resolution Egocentric Vision |
| topic | Robotics |
| url | https://arxiv.org/abs/2605.14106 |