Institutional Repository of Vilnius University
Neuroninių tinklų taikymas mikroplastikų nehomogeniniame fone segmentavimui /
Abstract
dc:descriptionMicroplastic pollution of oceans, rivers and lakes is a rapidly growing problem worldwide. The impact of these microscopic particles on humans and nature is not well understood, and there is a growing need for studies analysing water samples and their microplastic content, but currently there are no fast and cheap methods to count them, and new ways of counting microplastics are needed. In this bachelor thesis, based on the dataset provided by the FTMC and the analysis of existing machine learning segmentation models and computational methods, several convolutional neural network models are developed and compared, and the best parameters and methods for solving this problem are proposed. The developed model is able to accurately count microplastics in a given dataset, although all the highest segmentation metrics cannot be achieved, as the recall and F1 metrics decrease with increasing precision and the intersection over union metric, but the computational performance improves. Furthermore, experiments performed on colour space transformation and morphological operations on the dataset demonstrate the usefulness of these applications in improving the results. Such a model could help microplastics researchers to quickly and accurately determine the total number of microplastics in water samples.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Savickas, Audrius,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210647925&prefLang=en_US