Target Dataset

Description This dataset contains microscopy images of plastic fibers suspended in water, developed for a Master's thesis in Environmental Engineering at Instituto Superior Técnico (2026). The dataset was created to support semantic segmentation tasks using deep learning methods for microplastic detection and characterization. The repository includes: Training images Validation images Test images Corresponding segmentation masks (ground truth) The dataset was used to evaluate transfer learning strategies based on a U-Net architecture with a ResNet50 encoder pretrained on microscopy data. Related Resources GitHub repository containing preprocessing, training, and evaluation code: https://github.com/matildecrodriguess/microplastics-segmentation-in-water Intended Use This dataset is intended for: Semantic segmentation Microscopy image analysis Deep learning benchmarking Microplastic detection research Acknowledgements This work was carried out within the framework of the TWIST project – “Assessing the risks and benefits of using plastic fibres from the textile industry in mortars to address plastic waste” (2023.16868.ICDT) , funded by Fundação para a Ciência e a Tecnologia (FCT) , Portugal . https://doi.org/10.54499/2023.16868.ICDT.