Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition
Recent advances in big data systems and databases have made it possible to gather raw unlabeled data at unprecedented rates. However, labeling such data constitutes a costly and timely process. This is especially true for video data, and in particular for human activity recognition (HAR) tasks. For...
Main Authors: | , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-10-01
|
Series: | Technologies |
Subjects: | |
Online Access: | https://www.mdpi.com/2227-7080/8/4/55 |
_version_ | 1797550968251875328 |
---|---|
author | Evaggelos Spyrou Eirini Mathe Georgios Pikramenos Konstantinos Kechagias Phivos Mylonas |
author_facet | Evaggelos Spyrou Eirini Mathe Georgios Pikramenos Konstantinos Kechagias Phivos Mylonas |
author_sort | Evaggelos Spyrou |
collection | DOAJ |
description | Recent advances in big data systems and databases have made it possible to gather raw unlabeled data at unprecedented rates. However, labeling such data constitutes a costly and timely process. This is especially true for video data, and in particular for human activity recognition (HAR) tasks. For this reason, methods for reducing the need of labeled data for HAR applications have drawn significant attention from the research community. In particular, two popular approaches developed to address the above issue are <i>data augmentation</i> and <i>domain adaptation</i>. The former attempts to leverage problem-specific, hand-crafted data synthesizers to augment the training dataset with artificial labeled data instances. The latter attempts to extract knowledge from distinct but related supervised learning tasks for which labeled data is more abundant than the problem at hand. Both methods have been extensively studied and used successfully on various tasks, but a comprehensive comparison of the two has not been carried out in the context of video data HAR. In this work, we fill this gap by providing ample experimental results comparing data augmentation and domain adaptation techniques on a cross-viewpoint, human activity recognition task from pose information. |
first_indexed | 2024-03-10T15:37:59Z |
format | Article |
id | doaj.art-e921ef048e354b81b16bebbfc06f6b5e |
institution | Directory Open Access Journal |
issn | 2227-7080 |
language | English |
last_indexed | 2024-03-10T15:37:59Z |
publishDate | 2020-10-01 |
publisher | MDPI AG |
record_format | Article |
series | Technologies |
spelling | doaj.art-e921ef048e354b81b16bebbfc06f6b5e2023-11-20T17:08:57ZengMDPI AGTechnologies2227-70802020-10-01845510.3390/technologies8040055Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity RecognitionEvaggelos Spyrou0Eirini Mathe1Georgios Pikramenos2Konstantinos Kechagias3Phivos Mylonas4Department of Computer Science and Telecommunications, University of Thessaly, 35131 Lamia, GreeceInstitute of Informatics and Telecommunications, National Center for Scientific Research—“Demokritos”, 15310 Athens, GreeceInstitute of Informatics and Telecommunications, National Center for Scientific Research—“Demokritos”, 15310 Athens, GreeceDepartment of Informatics and Telecommunications, University of Athens, 15784 Athens, GreeceDepartment of Informatics, Ionian University, 49100 Corfu, GreeceRecent advances in big data systems and databases have made it possible to gather raw unlabeled data at unprecedented rates. However, labeling such data constitutes a costly and timely process. This is especially true for video data, and in particular for human activity recognition (HAR) tasks. For this reason, methods for reducing the need of labeled data for HAR applications have drawn significant attention from the research community. In particular, two popular approaches developed to address the above issue are <i>data augmentation</i> and <i>domain adaptation</i>. The former attempts to leverage problem-specific, hand-crafted data synthesizers to augment the training dataset with artificial labeled data instances. The latter attempts to extract knowledge from distinct but related supervised learning tasks for which labeled data is more abundant than the problem at hand. Both methods have been extensively studied and used successfully on various tasks, but a comprehensive comparison of the two has not been carried out in the context of video data HAR. In this work, we fill this gap by providing ample experimental results comparing data augmentation and domain adaptation techniques on a cross-viewpoint, human activity recognition task from pose information.https://www.mdpi.com/2227-7080/8/4/55human activity recognitiondata augmentationdata adaptationactivities of daily living |
spellingShingle | Evaggelos Spyrou Eirini Mathe Georgios Pikramenos Konstantinos Kechagias Phivos Mylonas Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition Technologies human activity recognition data augmentation data adaptation activities of daily living |
title | Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition |
title_full | Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition |
title_fullStr | Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition |
title_full_unstemmed | Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition |
title_short | Data Augmentation vs. Domain Adaptation—A Case Study in Human Activity Recognition |
title_sort | data augmentation vs domain adaptation a case study in human activity recognition |
topic | human activity recognition data augmentation data adaptation activities of daily living |
url | https://www.mdpi.com/2227-7080/8/4/55 |
work_keys_str_mv | AT evaggelosspyrou dataaugmentationvsdomainadaptationacasestudyinhumanactivityrecognition AT eirinimathe dataaugmentationvsdomainadaptationacasestudyinhumanactivityrecognition AT georgiospikramenos dataaugmentationvsdomainadaptationacasestudyinhumanactivityrecognition AT konstantinoskechagias dataaugmentationvsdomainadaptationacasestudyinhumanactivityrecognition AT phivosmylonas dataaugmentationvsdomainadaptationacasestudyinhumanactivityrecognition |