Multistream ValidNet: Improving 6D Object Pose Estimation by Automatic Multistream Validation

This work presents a novel approach to improve the results of pose estimation\nby detecting and distinguishing between the occurrence of True and False\nPositive results. It achieves this by training a binary classifier on the\noutput of an arbitrary pose estimation algorithm, and returns a binary label\nindicating the validity of the result. We demonstrate that our approach\nimproves upon a state-of-the-art pose estimation result on the Sil\\'eane\ndataset, outperforming a variation of the alternative CullNet method by 4.15%\nin average class accuracy and 0.73% in overall accuracy at validation. Applying\nour method can also improve the pose estimation average precision results of\nOp-Net by 6.06% on average.\n