Related Experiment Video
Updated: Aug 5, 2026
10:16
Protocol for Microplastics Sampling on the Sea Surface and Sample Analysis
Published on: December 16, 2016
Can AI Reliably Identify Marine Microplastics in Wildlife? Assessing Multi-Modal Foundation Models for Polymer
Gabriela Fernandez1, Domenico Vito1, Siddharth Suresh-Babu1
1Metabolism of Cities Living Lab, Center for Human Dynamics in the Mobile Age, San Diego State University, 5500 Campanile Drive, San Diego, CA 92185, USA.
Summary
General-purpose AI models can classify microplastics from coastal images with minimal resources. This study shows their potential as accessible tools for marine pollution monitoring.
Area of Science:
- Environmental Science
- Artificial Intelligence
- Marine Biology
Background:
- Existing AI microplastic studies often use specialized models, requiring significant resources.
- Accessible, low-cost tools are needed for effective marine pollution monitoring.
Purpose of the Study:
- To evaluate general-purpose multimodal foundation models for microplastic classification.
- To assess their utility as accessible, low-cost tools for field-based monitoring.
- To analyze performance under ecologically realistic conditions.
Main Methods:
- A dataset of 1080 high-resolution images of plastic debris from Tunisia was created.
- Images were annotated with pixel-level segmentation masks and class labels.
- Multimodal large language models (LLMs) were evaluated using spatial accuracy (mIoU) and classification performance (F1 scores).
Main Results:
- Multimodal foundation models successfully distinguished plastic fragments from natural backgrounds.
- Performance varied across different polymer types and under environmental perturbations.
- A composite scoring framework prioritized ecological relevance over computational metrics.
Conclusions:
- General-purpose multimodal foundation models show promise as accessible tools for microplastic analysis.
- Further research is needed to optimize performance across diverse environmental conditions.
- Findings inform AI applications in marine pollution monitoring and One Health initiatives.
