VeREFINE: Integrating Object Pose Verification with Iterative Physics-guided Refinement

Precise object pose estimation for robotics applications and augmented reality relies on final refinement and verification steps. However, interactions between objects and interactions with the supporting structures in the observed scene are typically not considered. In this work, we propose to integrate scene-level hypotheses verification with object-level object pose refinement guided by physics simulation. This allows the physical plausibility of individual object pose estimates and the stability of the estimated scene to be consider in a unified search-based optimization. The proposed method is able to adapt to scenes of multiple objects and efficiently focuses on refining the most promising object poses in multi-hypotheses scenarios. We call this integrated approach VeREFINE and evaluate it on two datasets with varying scene complexity. The generality of the approach is shown by using two different pose estimators and two different baseline refiners. Results show improvements over all baselines and on all datasets with the inclusion of our proposed VeREFINE approach.