Few-shot learners aim to recognize new object classes based on a small number of labeled training examples. To prevent overfitting, state-of-the-art few-shot learners use meta-learning on convolutional-network features and perform classification using a nearest-neighbor classifier. This paper studies the accuracy of nearest-neighbor baselines without meta-learning. Surprisingly, we find simple feature transformations suffice to obtain competitive few-shot learning accuracies. For example, we find that a nearest-neighbor classifier used in combination with mean-subtraction and L2-normalization outperforms prior results in three out of five settings on the miniImageNet dataset.
@article{arxiv.1911.04623,
title = {SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning},
author = {Yan Wang and Wei-Lun Chao and Kilian Q. Weinberger and Laurens van der Maaten},
journal= {arXiv preprint arXiv:1911.04623},
year = {2019}
}