Producing plankton classifiers that are robust to dataset shift

IF 2.1 3区 地球科学 Q2 LIMNOLOGY Limnology and Oceanography: Methods Pub Date : 2024-11-27 DOI:10.1002/lom3.10659
C. Chen, S. P. Kyathanahally, M. Reyes, S. Merkli, E. Merz, E. Francazi, M. Hoege, F. Pomati, M. Baity-Jesi
{"title":"Producing plankton classifiers that are robust to dataset shift","authors":"C. Chen,&nbsp;S. P. Kyathanahally,&nbsp;M. Reyes,&nbsp;S. Merkli,&nbsp;E. Merz,&nbsp;E. Francazi,&nbsp;M. Hoege,&nbsp;F. Pomati,&nbsp;M. Baity-Jesi","doi":"10.1002/lom3.10659","DOIUrl":null,"url":null,"abstract":"<p>Modern plankton high-throughput monitoring relies on deep learning classifiers for species recognition in water ecosystems. Despite satisfactory nominal performances, a significant challenge arises from dataset shift, which causes performances to drop during deployment. In our study, we integrate the ZooLake dataset, which consists of dark-field images of lake plankton (Kyathanahally et al. 2021a), with manually annotated images from 10 independent days of deployment, serving as <i>test cells</i> to benchmark out-of-dataset (OOD) performances. Our analysis reveals instances where classifiers, initially performing well in in-dataset conditions, encounter notable failures in practical scenarios. For example, a MobileNet with a 92% nominal test accuracy shows a 77% OOD accuracy. We systematically investigate conditions leading to OOD performance drops and propose a preemptive assessment method to identify potential pitfalls when classifying new data, and pinpoint features in OOD images that adversely impact classification. We present a three-step pipeline: (i) identifying OOD degradation compared to nominal test performance, (ii) conducting a diagnostic analysis of degradation causes, and (iii) providing solutions. We find that ensembles of BEiT vision transformers, with targeted augmentations addressing OOD robustness, geometric ensembling, and rotation-based test-time augmentation, constitute the most robust model, which we call <i>BEsT</i>. It achieves an 83% OOD accuracy, with errors concentrated on container classes. Moreover, it exhibits lower sensitivity to dataset shift, and reproduces well the plankton abundances. Our proposed pipeline is applicable to generic plankton classifiers, contingent on the availability of suitable test cells. By identifying critical shortcomings and offering practical procedures to fortify models against dataset shift, our study contributes to the development of more reliable plankton classification technologies.</p>","PeriodicalId":18145,"journal":{"name":"Limnology and Oceanography: Methods","volume":"23 1","pages":"39-66"},"PeriodicalIF":2.1000,"publicationDate":"2024-11-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Limnology and Oceanography: Methods","FirstCategoryId":"89","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/lom3.10659","RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"LIMNOLOGY","Score":null,"Total":0}
引用次数: 0

Abstract

Modern plankton high-throughput monitoring relies on deep learning classifiers for species recognition in water ecosystems. Despite satisfactory nominal performances, a significant challenge arises from dataset shift, which causes performances to drop during deployment. In our study, we integrate the ZooLake dataset, which consists of dark-field images of lake plankton (Kyathanahally et al. 2021a), with manually annotated images from 10 independent days of deployment, serving as test cells to benchmark out-of-dataset (OOD) performances. Our analysis reveals instances where classifiers, initially performing well in in-dataset conditions, encounter notable failures in practical scenarios. For example, a MobileNet with a 92% nominal test accuracy shows a 77% OOD accuracy. We systematically investigate conditions leading to OOD performance drops and propose a preemptive assessment method to identify potential pitfalls when classifying new data, and pinpoint features in OOD images that adversely impact classification. We present a three-step pipeline: (i) identifying OOD degradation compared to nominal test performance, (ii) conducting a diagnostic analysis of degradation causes, and (iii) providing solutions. We find that ensembles of BEiT vision transformers, with targeted augmentations addressing OOD robustness, geometric ensembling, and rotation-based test-time augmentation, constitute the most robust model, which we call BEsT. It achieves an 83% OOD accuracy, with errors concentrated on container classes. Moreover, it exhibits lower sensitivity to dataset shift, and reproduces well the plankton abundances. Our proposed pipeline is applicable to generic plankton classifiers, contingent on the availability of suitable test cells. By identifying critical shortcomings and offering practical procedures to fortify models against dataset shift, our study contributes to the development of more reliable plankton classification technologies.

查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
求助全文
约1分钟内获得全文 去求助
来源期刊
CiteScore
4.80
自引率
3.70%
发文量
56
审稿时长
3 months
期刊介绍: Limnology and Oceanography: Methods (ISSN 1541-5856) is a companion to ASLO''s top-rated journal Limnology and Oceanography, and articles are held to the same high standards. In order to provide the most rapid publication consistent with high standards, Limnology and Oceanography: Methods appears in electronic format only, and the entire submission and review system is online. Articles are posted as soon as they are accepted and formatted for publication. Limnology and Oceanography: Methods will consider manuscripts whose primary focus is methodological, and that deal with problems in the aquatic sciences. Manuscripts may present new measurement equipment, techniques for analyzing observations or samples, methods for understanding and interpreting information, analyses of metadata to examine the effectiveness of approaches, invited and contributed reviews and syntheses, and techniques for communicating and teaching in the aquatic sciences.
期刊最新文献
Issue Information Issue Information Deep learning-based characterization of underwater methane bubbles using simple dual camera platform Underwater Zooplankton Enhancement Light Array (UZELA): A technology solution to enhance zooplankton abundance and coral feeding in bleached and non-bleached corals Inferring microalgae density and net ecosystem production on soft sediments using infrared imaging
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 联系我们:info@booksci.cn
Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。
Copyright © 2023 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1