Segmentación automática de células de Allium cepa por métodos no supervisados de visión artificial
Date
2025-10-01
Journal Title
Journal ISSN
Volume Title
Publisher
Facultad Regional Buenos Aires
Abstract
El presente estudio aborda la segmentación e identificación de células mediante visión artificial en el contexto de la prueba de Allium cepa para el monitoreo de efluentes líquidos. Se propone un método de detección generalista basado en el modelo SAM (Segment Anything Model), complementado con algoritmos no supervisados para separar las regiones de ruido de las regiones con células, y posteriormente aplicar un clasificador para identificar las instancias celulares. El modelo fue evaluado utilizando imágenes proporcionadas por especialistas del Instituto Nacional del Agua, así como con un conjunto de validación externo, obteniendo exactitudes del ~96 % y ~95 %, respectivamente.
This study addresses the automation of cell segmentation and identification using computer vision techniques in the context of the Allium cepa test for monitoring liquid effluents. The implementation of a generalist detection method based on the model SAM (Segment Anything Model) is proposed, complemented by unsupervised algorithms to separate noise regions from cells, followed by the application of a classifier to identify cellular instances. The model's performance is evaluated using images provided by specialists from the National Water Institute, as well as with an independent external validation set, achieving accuracies of ~96% and ~95%, respectively.
This study addresses the automation of cell segmentation and identification using computer vision techniques in the context of the Allium cepa test for monitoring liquid effluents. The implementation of a generalist detection method based on the model SAM (Segment Anything Model) is proposed, complemented by unsupervised algorithms to separate noise regions from cells, followed by the application of a classifier to identify cellular instances. The model's performance is evaluated using images provided by specialists from the National Water Institute, as well as with an independent external validation set, achieving accuracies of ~96% and ~95%, respectively.
Description
Keywords
Visión artificial, Aprendizaje no supervisado, Allium cepa, Medio ambiente, Ecotoxicología, Artificial vision, Not supervised learning, Allium cepa, Environment, Ecotoxicology
Citation
Proyecciones, Vol.23 Nº 2
Collections
Endorsement
Review
Supplemented By
Referenced By
Creative Commons license
Except where otherwised noted, this item's license is described as info:eu-repo/semantics/openAccess