Sensors Journal

Deep Bayesian-Assisted Keypoint Detection for Pose Estimation in Assembly Automation

Quick Summary

  • Congratulations to Nick Shi and Amin Ghafourian and our collaborators at Ford Motor Company for publishing their work in Sensors.

Our research on keypoint-based object localization for assembly automation published in Sensors. 

Summary: Assembly processes are in dire need of innovative and simple techniques to streamline automation. In this paper we discuss our approach to data-efficient learning for assembly automation in which we rely on the availability of geometrical characteristics of mechanical components on the production floors. We formulate pose estimation in the context of Bayesian statistics and improve data-efficiency of a deep network trained on limited data.