Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis
Novel neural network models that can handle complex tasks with fewer examples than before are being developed for a wide range of applications. In some fields, even the creation of a few labels is a laborious task and impractical, especially for data that require more than a few seconds to generate...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-04-01
|
| Series: | Algorithms |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1999-4893/16/4/205 |
| _version_ | 1797606749052600320 |
|---|---|
| author | Dominik Stallmann Barbara Hammer |
| author_facet | Dominik Stallmann Barbara Hammer |
| author_sort | Dominik Stallmann |
| collection | DOAJ |
| description | Novel neural network models that can handle complex tasks with fewer examples than before are being developed for a wide range of applications. In some fields, even the creation of a few labels is a laborious task and impractical, especially for data that require more than a few seconds to generate each label. In the biotechnological domain, cell cultivation experiments are usually done by varying the circumstances of the experiments, seldom in such a way that hand-labeled data of one experiment cannot be used in others. In this field, exact cell counts are required for analysis, and even by modern standards, semi-supervised models typically need hundreds of labels to achieve acceptable accuracy on this task, while classical image processing yields unsatisfactory results. We research whether an unsupervised learning scheme is able to accomplish this task without manual labeling of the given data. We present a VAE-based Siamese architecture that is expanded in a cyclic fashion to allow the use of labeled synthetic data. In particular, we focus on generating pseudo-natural images from synthetic images for which the target variable is known to mimic the existence of labeled natural data. We show that this learning scheme provides reliable estimates for multiple microscopy technologies and for unseen data sets without manual labeling. We provide the source code as well as the data we use. The code package is open source and free to use (MIT licensed). |
| first_indexed | 2024-03-11T05:19:28Z |
| format | Article |
| id | doaj.art-f536869cd834416da88f7db49dfa4408 |
| institution | Directory Open Access Journal |
| issn | 1999-4893 |
| language | English |
| last_indexed | 2024-03-11T05:19:28Z |
| publishDate | 2023-04-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Algorithms |
| spelling | doaj.art-f536869cd834416da88f7db49dfa44082023-11-17T17:59:15ZengMDPI AGAlgorithms1999-48932023-04-0116420510.3390/a16040205Unsupervised Cyclic Siamese Networks Automating Cell Imagery AnalysisDominik Stallmann0Barbara Hammer1Faculty of Technology, University of Bielefeld, Universitätsstraße 25, 33615 Bielefeld, GermanyFaculty of Technology, University of Bielefeld, Universitätsstraße 25, 33615 Bielefeld, GermanyNovel neural network models that can handle complex tasks with fewer examples than before are being developed for a wide range of applications. In some fields, even the creation of a few labels is a laborious task and impractical, especially for data that require more than a few seconds to generate each label. In the biotechnological domain, cell cultivation experiments are usually done by varying the circumstances of the experiments, seldom in such a way that hand-labeled data of one experiment cannot be used in others. In this field, exact cell counts are required for analysis, and even by modern standards, semi-supervised models typically need hundreds of labels to achieve acceptable accuracy on this task, while classical image processing yields unsatisfactory results. We research whether an unsupervised learning scheme is able to accomplish this task without manual labeling of the given data. We present a VAE-based Siamese architecture that is expanded in a cyclic fashion to allow the use of labeled synthetic data. In particular, we focus on generating pseudo-natural images from synthetic images for which the target variable is known to mimic the existence of labeled natural data. We show that this learning scheme provides reliable estimates for multiple microscopy technologies and for unseen data sets without manual labeling. We provide the source code as well as the data we use. The code package is open source and free to use (MIT licensed).https://www.mdpi.com/1999-4893/16/4/205Siamese networkssynthetic datacyclic learningunsupervised learningdeep learningdata augmentation |
| spellingShingle | Dominik Stallmann Barbara Hammer Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis Algorithms Siamese networks synthetic data cyclic learning unsupervised learning deep learning data augmentation |
| title | Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis |
| title_full | Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis |
| title_fullStr | Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis |
| title_full_unstemmed | Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis |
| title_short | Unsupervised Cyclic Siamese Networks Automating Cell Imagery Analysis |
| title_sort | unsupervised cyclic siamese networks automating cell imagery analysis |
| topic | Siamese networks synthetic data cyclic learning unsupervised learning deep learning data augmentation |
| url | https://www.mdpi.com/1999-4893/16/4/205 |
| work_keys_str_mv | AT dominikstallmann unsupervisedcyclicsiamesenetworksautomatingcellimageryanalysis AT barbarahammer unsupervisedcyclicsiamesenetworksautomatingcellimageryanalysis |