One-Shot Object Localization Using Learnt Visual Cues via Siamese Networks
This work tackles the problem of one-shot object localization for robots operating in novel and unstructured environments, which is an incremental step towards more adaptable robotic systems.
This paper addresses the problem of localizing novel, previously unseen objects in new environments using a visual cue. The authors demonstrate that a simulated robot can pick-and-place novel objects specified by a laser pointer, and evaluate their approach on modified Omniglot and a small toy dataset.
A robot that can operate in novel and unstructured environments must be capable of recognizing new, previously unseen, objects. In this work, a visual cue is used to specify a novel object of interest which must be localized in new environments. An end-to-end neural network equipped with a Siamese network is used to learn the cue, infer the object of interest, and then to localize it in new environments. We show that a simulated robot can pick-and-place novel objects pointed to by a laser pointer. We also evaluate the performance of the proposed approach on a dataset derived from the Omniglot handwritten character dataset and on a small dataset of toys.