Deep-MAP is a Google Colaboratory-based tool for automated microplastic classification and quantification from microscope images. It uses a deep learning model to identify microplastic count, types, areas, and colors, helping researchers overcome manual analysis limitations.

About Deep-MAP

Deep-MAP is a deep learning-based tool for automated microplastic (MP) classification and quantification from microscope images. It addresses the time-consuming, labor-intensive, and biased nature of manual analysis. Y. Okumura and a team from Nagoya City University, Japan, and Universitas Brawijaya, Indonesia, developed Deep-MAP to provide high-precision automated analysis for environmental scientists and researchers, eliminating the need for specialized deep learning expertise.

The tool employs an end-to-end segmentation model, YOLOv8m-seg, for simultaneous MP detection, classification, and shape identification. Users upload microscope images to the Google Colaboratory interface. Deep-MAP then automatically generates aggregated outputs including MP count, types (Fiber, Fragment, Film, Foam, Pellet), areas, and colors. This approach is optimized for standard optical microscope images and effectively detects small microplastics (SMPs).

Key Features of Deep-MAP

• Automated classification of microplastic particles into five predefined shapes: Fiber, Fragment, Film, Foam, and Pellet.
• Quantification of microplastic count, individual particle areas, and colors directly from uploaded microscope images.
• Powered by a high-performance YOLOv8m-seg deep learning model, validated for instance segmentation tasks.
• User-friendly interface implemented on Google Colaboratory, minimizing the need for advanced programming or deep learning skills.
• Optimized for analysis of standard optical microscope images, including effective detection of small microplastics (SMPs).

Development and Validation

A collaborative team from Nagoya City University, Japan (Y. Okumura, A. Fadhilah, S. Kidou), and Universitas Brawijaya, Indonesia (R. Haribowo), developed Deep-MAP. Their peer-reviewed publication details the tool’s development and validation: Okumura et al., 2026, IOP Conf. Ser.: Earth Environ. Sci.

The underlying deep learning model achieved strong validation metrics, with a Mask mAP@0.50–0.95 of 0.555 and Mask mAP@0.50 of 0.873. It outperformed alternative two-stage and pre-processing models in systematic evaluations. Researchers trained the model on a diverse dataset of microscope images, including freshwater samples from Indonesian rivers and those near Nagoya City, Japan. This dataset featured polygon annotations for five microplastic classes, employing selective data augmentation and iterative stratification for robust training.

Access and Data Availability

Deep-MAP is accessible as an open-access tool through Google Colaboratory. Users can directly interact with the platform via its dedicated notebook: Deep-MAP Google Colaboratory Link. The associated research paper detailing the model, methodology, and validation is published under the Creative Commons Attribution 4.0 licence, ensuring open access to its scientific content.

The tool primarily functions as an image analysis service, where users upload their microscope images for processing. It provides aggregated analytical outputs such as particle counts, classified types, measured areas, and identified colors. The raw training dataset itself is described within the paper but is not explicitly provided as a downloadable component of the Colaboratory tool.

Why This Resource Belongs on Plastiverse

Deep-MAP provides a valuable solution for microplastic researchers focused on morphological characterization from optical microscope images. It directly supports automated identification and quantification, reducing manual effort and bias. By offering deep learning through Google Colaboratory, it closes a critical gap for researchers without specialized programming skills. This tool’s ability to classify microplastic shapes and detect small microplastics (SMPs) in diverse freshwater matrices enhances the precision and reproducibility of environmental monitoring, contributing essential data for understanding pollution states, complementing resources like the Interactive Mapping Application for marine microplastics or characterization insights from LitChemPlast – Chemicals in Plastics Measurement Database.