Deep Learning Model for Microplastic Categorization

This document is a research article that presents a deep learning segmentation model designed for the categorization of microplastic debris. The study outlines the critical need for efficient monitoring of microplastics in marine environments, emphasizing the limitations of traditional manual sampling methods, which are time-consuming and costly. The proposed workflow integrates advanced image processing techniques to facilitate both quantitative and qualitative assessments of microplastics ranging from 0.05 cm to 0.5 cm in size. The model demonstrates a high accuracy of 96% in identifying microplastic particles, providing a more robust and standardized alternative to manual counting. Additionally, the authors have made the models and datasets open-source, along with a user-friendly web interface for annotating new images. This initiative aims to support researchers in the field of plastic pollution by enabling standardized and comparable analysis of microplastic samples across various environmental compartments.