Publication details

COMPYDA: An online tool for verifying the similarity of image datasets

Authors

NEČASOVÁ Tereza MÚČKA Daniel SVOBODA David

Year of publication 2024
Type Article in Proceedings
Conference 2024 IEEE International Symposium on Biomedical Imaging (ISBI)
MU Faculty or unit

Faculty of Informatics

Citation
web https://ieeexplore.ieee.org/document/10635415
Doi http://dx.doi.org/10.1109/ISBI56570.2024.10635415
Keywords Web-service;Statistics;Data validation;Image descriptors;Privacy;Augmented data
Description Nowadays, when the vast majority of biomedical research relies on machine learning methods, paying attention to the meaningfulness of the data we work with is crucial. This is especially true if the dataset is scarce and we are required to use various augmentation techniques to enlarge training sets. The additional data are, however, not guaranteed to have the same characteristics as the original data, and therefore, the augmented set may be inconsistent. This can subsequently lead to incorrect training of biomedical image analysis methods, which may result in biased classification, detection, segmentation, or tracking results. In this paper, we present an online tool called COMPYDA, that allows users to easily assess the similarity of a pair of datasets using well-founded, commonly used statistic methods. COMPYDA guides users through univariate and multivariate analyses and helps them understand and explain dataset differences to ascertain a compatible dataset for further training. The tool is available at: https://cbia.fi.muni.cz/compyda/
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.

More info