Investigations on Output Parameterizations of Neural Networks for Single Shot 6D Object Pose Estimation

Single shot approaches have demonstrated tremendous success on various\ncomputer vision tasks. Finding good parameterizations for 6D object pose\nestimation remains an open challenge. In this work, we propose different novel\nparameterizations for the output of the neural network for single shot 6D\nobject pose estimation. Our learning-based approach achieves state-of-the-art\nperformance on two public benchmark datasets. Furthermore, we demonstrate that\nthe pose estimates can be used for real-world robotic grasping tasks without\nadditional ICP refinement.\n