Using Unlabeled Data for Increasing Low-Shot Classification Accuracy of Relevant and Open-Set Irrelevant Images

TL;DR

This paper introduces an open-set low-shot classifier that leverages unlabeled irrelevant images during training to improve identification of relevant, irrelevant, and unseen irrelevant images in autonomous vehicle tasks.

Contribution

It presents a novel low-shot classifier that uses unlabeled irrelevant images during training, enhancing open-set recognition with minimal labeled data.

Findings

Effective in identifying relevant and irrelevant images

Capable of recognizing unseen irrelevant categories

Compatible with pre-trained CNN feature extractors

Abstract

In search, exploration, and reconnaissance tasks performed with autonomous ground vehicles, an image classification capability is needed for specifically identifying targeted objects (relevant classes) and at the same time recognize when a candidate image does not belong to anyone of the relevant classes (irrelevant images). In this paper, we present an open-set low-shot classifier that uses, during its training, a modest number (less than 40) of labeled images for each relevant class, and unlabeled irrelevant images that are randomly selected at each epoch of the training process. The new classifier is capable of identifying images from the relevant classes, determining when a candidate image is irrelevant, and it can further recognize categories of irrelevant images that were not included in the training (unseen). The proposed low-shot classifier can be attached as a top layer to any pre-trained feature extractor when constructing a Convolutional Neural Network.