Transformation-Based Data Synthesis for Limited Sample Scenario

We consider a challenging learning scenario where neither pretext training nor auxiliary data are available except for small training samples. We call this a transfer-free scenario where we cannot access any transferable knowledge or data. Our proposal for resolving this issue is to learn a pair-wis...

Full description

Saved in:
Bibliographic Details
Main Authors: Chang-Hwa Lee, Sang Wan Lee
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10781377/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846123603992510464
author Chang-Hwa Lee
Sang Wan Lee
author_facet Chang-Hwa Lee
Sang Wan Lee
author_sort Chang-Hwa Lee
collection DOAJ
description We consider a challenging learning scenario where neither pretext training nor auxiliary data are available except for small training samples. We call this a transfer-free scenario where we cannot access any transferable knowledge or data. Our proposal for resolving this issue is to learn a pair-wise transformation function (e.g., spatial or appearance) between given samples. This simple setting yields two practical advantages. The training objective can be defined as a simple reconstruction loss, and data can be synthesized by merely manipulating or sampling the learned transformations. However, the limitation of previous transformation methods lies in a strong assumption that all images should be transformable to each other, i.e., all-to-all transformable. To relax this constraint, we propose a novel concept called ‘template,’ designed to be transformable to any other data, i.e., “template-to-all” transformable. A range of experiments on the transfer-free scenarios confirms that our model successfully learns transformation and synthesizes new data from minimal training data (less than five or ten for each class). The subsequent data augmentation experiments showed significantly improved classification performance.
format Article
id doaj-art-058a4a6afc79497587b09dcb73ace66e
institution Kabale University
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-058a4a6afc79497587b09dcb73ace66e2024-12-14T00:00:55ZengIEEEIEEE Access2169-35362024-01-011218484118485210.1109/ACCESS.2024.351253810781377Transformation-Based Data Synthesis for Limited Sample ScenarioChang-Hwa Lee0https://orcid.org/0000-0001-5961-8284Sang Wan Lee1https://orcid.org/0000-0001-6266-9613Brain and Cognitive Engineering Program, Korea Advanced Institute of Science Technology (KAIST), Daejeon, South KoreaDepartment of Brain and Cognitive Sciences, Department of Bio and Brain Engineering, Kim Jaechul Graduate School of AI, Center for Neuroscience-inspired AI, Korea Advanced Institute of Science Technology (KAIST), Daejeon, South KoreaWe consider a challenging learning scenario where neither pretext training nor auxiliary data are available except for small training samples. We call this a transfer-free scenario where we cannot access any transferable knowledge or data. Our proposal for resolving this issue is to learn a pair-wise transformation function (e.g., spatial or appearance) between given samples. This simple setting yields two practical advantages. The training objective can be defined as a simple reconstruction loss, and data can be synthesized by merely manipulating or sampling the learned transformations. However, the limitation of previous transformation methods lies in a strong assumption that all images should be transformable to each other, i.e., all-to-all transformable. To relax this constraint, we propose a novel concept called ‘template,’ designed to be transformable to any other data, i.e., “template-to-all” transformable. A range of experiments on the transfer-free scenarios confirms that our model successfully learns transformation and synthesizes new data from minimal training data (less than five or ten for each class). The subsequent data augmentation experiments showed significantly improved classification performance.https://ieeexplore.ieee.org/document/10781377/Image synthesissmall sample classificationcomputer visiondeep learning
spellingShingle Chang-Hwa Lee
Sang Wan Lee
Transformation-Based Data Synthesis for Limited Sample Scenario
IEEE Access
Image synthesis
small sample classification
computer vision
deep learning
title Transformation-Based Data Synthesis for Limited Sample Scenario
title_full Transformation-Based Data Synthesis for Limited Sample Scenario
title_fullStr Transformation-Based Data Synthesis for Limited Sample Scenario
title_full_unstemmed Transformation-Based Data Synthesis for Limited Sample Scenario
title_short Transformation-Based Data Synthesis for Limited Sample Scenario
title_sort transformation based data synthesis for limited sample scenario
topic Image synthesis
small sample classification
computer vision
deep learning
url https://ieeexplore.ieee.org/document/10781377/
work_keys_str_mv AT changhwalee transformationbaseddatasynthesisforlimitedsamplescenario
AT sangwanlee transformationbaseddatasynthesisforlimitedsamplescenario