Overall Positive Prototype for Few-Shot Open-Set Recognition

Liang-Yu Sun and Wei-Ta Chu

Multimedia and Computer Vision Laboratory
Dept. of Computer Science and Information Engineering
National Cheng Kung University

1. Introduction

Few-shot open-set recognition (FSOR) is the task of recognizing samples in known classes with a limited number of annotated instances while also de- tecting samples that do not belong to any known class. This is a challenging problem because the models must learn to generalize from a small number of labeled samples and distinguish them from an unlimited number of potential negative examples. In this paper, we propose a novel approach called overall positive prototype to effectively improve performance. Conceptually, nega-tive samples would distribute throughout the feature space and are hard to be described. From the opposite viewpoint, we propose to construct an overall positive prototype that acts as a cohesive representation for positive sam-ples that distribute in a relatively smaller neighborhood. By measuring the distance between a query sample and the overall positive prototype, we can effectively classify it as either positive or negative. We show that this simple yet innovative approach provides the state-of-the-art FSOR performance in terms of accuracy and AUROC.

2. GitHub Link

3. Citation

Please cite our work if you utilize this dataset.

Liang-Yu Sun and Wei-Ta Chu, "Overall Positive Prototype for Few-Shot Open-Set Recognition," Pattern Recognition, vol. 151, pp. 110400, 2024.
https://doi.org/10.1016/j.patcog.2024.110400

Any problem please contact .

[Main Page]

Last Updated: May 12, 2024