Frozen Feature Augmentation for Few-Shot Image Classification

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bär, Andreas, Houlsby, Neil, Dehghani, Mostafa, Kumar, Manoj
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913446377816064
author Bär, Andreas
Houlsby, Neil
Dehghani, Mostafa
Kumar, Manoj
author_facet Bär, Andreas
Houlsby, Neil
Dehghani, Mostafa
Kumar, Manoj
contents Training a linear classifier or lightweight model on top of pretrained vision model outputs, so-called 'frozen features', leads to impressive performance on a number of downstream few-shot tasks. Currently, frozen features are not modified during training. On the other hand, when networks are trained directly on images, data augmentation is a standard recipe that improves performance with no substantial overhead. In this paper, we conduct an extensive pilot study on few-shot image classification that explores applying data augmentations in the frozen feature space, dubbed 'frozen feature augmentation (FroFA)', covering twenty augmentations in total. Our study demonstrates that adopting a deceptively simple pointwise FroFA, such as brightness, can improve few-shot performance consistently across three network architectures, three large pretraining datasets, and eight transfer datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2403_10519
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Frozen Feature Augmentation for Few-Shot Image Classification
Bär, Andreas
Houlsby, Neil
Dehghani, Mostafa
Kumar, Manoj
Computer Vision and Pattern Recognition
Training a linear classifier or lightweight model on top of pretrained vision model outputs, so-called 'frozen features', leads to impressive performance on a number of downstream few-shot tasks. Currently, frozen features are not modified during training. On the other hand, when networks are trained directly on images, data augmentation is a standard recipe that improves performance with no substantial overhead. In this paper, we conduct an extensive pilot study on few-shot image classification that explores applying data augmentations in the frozen feature space, dubbed 'frozen feature augmentation (FroFA)', covering twenty augmentations in total. Our study demonstrates that adopting a deceptively simple pointwise FroFA, such as brightness, can improve few-shot performance consistently across three network architectures, three large pretraining datasets, and eight transfer datasets.
title Frozen Feature Augmentation for Few-Shot Image Classification
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.10519