Optimization Beyond the Convolution: Generalizing Spatial Relations with End-to-End Metric Learning

Supplementary video for our paper `Optimization Beyond the Convolution: Generalizing Spatial Relations with End-to-End Metric Learning`. A preprint can by found here: https://arxiv.org/abs/1707.00893

To operate intelligently in domestic environments, robots require the ability to understand arbitrary spatial relations between objects and to generalize them to objects of varying sizes and shapes. In this work, we present a novel end-to-end approach utilizing neural networks to generalize spatial relations based on distance metric learning. Our network transforms spatial relations to a feature space that captures their similarities based on 3D point clouds of the objects and without prior semantic knowledge of the relations. It employs gradient-based optimization to compute object poses in order to imitate an arbitrary target relation by reducing the distance to it under the learned metric.

Submitted to CoRL 2017.

• Category

  • Science & Technology

• License

  • Standard YouTube License