Event
OMAR: A Shelf-Organization Pipeline for Open-World Multi-Attribute Reasoning
1 reports · 1 sources
The arXiv paper proposes OMAR, an open-world multi-attribute reasoning pipeline for scene-conditioned shelf organization that infers a shelf's latent organization scheme and ranks placement positions for new objects when the user provides no preference specification. The pipeline works as follows: a vision-language model first proposes object labels and attributes such as contents and packaging type, which are grounded to an instance-level 3D scene graph via segmentation, depth, and multi-view association; a large language model reasons over this representation and produces scene-consistent placement rankings; the rankings then guide geometric and kinematic feasibility checks to execute placement. The paper claims the method supports autonomous placement of previously unseen objects without predefined object-container mappings or category-specific retraining.
Report timeline
- OMAR: Open-World Multi-Attribute Reasoning for Scene-Conditioned Shelf Organization
An arXiv paper proposes OMAR, an open-world multi-attribute reasoning pipeline for inferring a shelf's latent organization and ranking placement positions for new objects without user preference instructions. A vision-language model first proposes object labels and attributes such as content and packaging type, which are grounded through segmentation, depth, and multi-view association; an LLM then reasons over an instance-level 3D scene graph to produce scene-consistent placement rankings, followed by geometric and kinematic feasibility checks before execution.
arXiv cs.RO ·