Augmenting citizen science with computer vision for fish monitoring

2026-03-25 · Source: MIT News - Artificial intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, medium

Summary

MIT Sea Grant, Woodwell Climate Research Center, and collaborators have demonstrated a deep learning-based system for automated fish monitoring, detailed in their March 25, 2026 publication. This system addresses the limitations of traditional visual counting and volunteer-based programs for river herring, which suffer from time, environmental, and labor constraints. The new method utilizes underwater video and computer vision, including object detection, tracking, and species classification, to provide continuous, high-resolution fish counts. Researchers collected videos from three Massachusetts rivers, manually labeled 59,850 frames across 1,435 clips for training, and validated the computer vision counts against human reviews and PIT tagging data. The system successfully counted 42,510 river herring in the 2024 Coonamessett River migration, revealing migration patterns like peak upstream movement at dawn and nocturnal downstream activity.

Key takeaway

For fisheries managers and conservation groups assessing aquatic populations, this computer vision system offers a scalable, cost-effective, and efficient solution for automated fish monitoring. You should consider integrating deep learning-based video analysis to supplement traditional methods, gaining higher-resolution, season-long counts and insights into migration behaviors. This approach can enhance population assessments and inform conservation strategies more effectively.

Key insights

Computer vision and deep learning can automate fish monitoring, improving efficiency and data quality for conservation.

Principles

• Diverse multi-site, multi-year data improves model reliability.
• Automated systems provide insights into migration behavior.

Method

An end-to-end pipeline involves in-field underwater cameras, manual video labeling with bounding boxes for fish tracking, and model training on diverse datasets for automated counting and behavior analysis.

In practice

• Use underwater cameras for continuous monitoring.
• Annotate videos with bounding boxes for training.
• Integrate citizen science for camera maintenance.

Topics

• Computer Vision
• Deep Learning
• Fish Monitoring
• Citizen Science
• Fisheries Management

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, AI Researcher, Machine Learning Engineer, Data Scientist

Related on AIssential

• Machine learning framework to predict global imperilment status of freshwater fish — An AI-based model predicts freshwater fish imperilment by analyzing diverse variables to enable proactive conservation.
• AI Can Help Track the World’s Shrinking Glaciers — AI models for glacier tracking can achieve human-comparable accuracy in new regions with minimal targeted data and environmental context.
• Visual intelligence for aquaculture: what AI can see that humans miss — AI-driven visual intelligence enhances aquaculture by continuously monitoring existing camera feeds for subtle behavioral changes.
• National-scale acoustic monitoring of avian biodiversity and migration — Passive acoustic monitoring offers a scalable, automated method for tracking avian migration and biodiversity across vast regions.
• Colour Extraction Pipeline for Odonates using Computer Vision — A computer vision pipeline segments Odonates and extracts color data to enable large-scale ecological correlation analysis.
• MIT Researcher Talks AI & Environmental Conservation — AI accelerates environmental conservation by automating data analysis and optimizing model selection for diverse ecological challenges.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT News - Artificial intelligence.