Automated Monitoring of Insects (AMI)

As part of a global AMI consortium, Mila is helping scientists study insect populations worldwide, informing responses to climate and biodiversity crises.

Background

Insects account for about half of all living species on Earth, and are crucial to everything from pollination to soil health. Without a diverse insect population, many human activities would be in serious jeopardy. Sadly, due to a number of factors, including climate change, insect populations are in serious trouble.

Initiated in Summer 2020 and now in use on three continents, the AMI project studies insect biodiversity through cutting-edge technologies. As part of this international consortium, Mila is helping to revolutionize the way insects are collected, identified and monitored through its work on the AMI Data Companion, machine-learning algorithms, out-of-distribution detection techniques, and more.

Objectives

There are an estimated 10 million insect species on Earth — and approximately 1.4 billion insects for every human being (Royal Entomological Society, 2024) — making it challenging to properly collect and study insects using traditional methods. Further complicating this important research is a lack of trained experts worldwide.

The AMI project is helping to change that. Through technologies such as high-resolution cameras, low-cost sensors, and AI-based processing methods, insect monitoring has become less labour-intensive, more user-friendly, and more effective at determining the health of critical insect populations across the globe.

About the Project

How does it work?

Using insect cameras as a starting point, the machine-learning open-source software developed by Mila is helping researchers to assess and use what the cameras capture.

An object detector begins by visually differentiating each insect in an image. An image classifier then separates moths (the current focus of the project) from other insects, and subjects the moths to a species classifier. Finally, a tracking algorithm counts the number of individual moths across frames, building up a picture of insect health in a particular area.

To make machine learning tools easier for ecologists to use, the AMI Data Companion software helps researchers analyze their camera-trap data. Also in the pipeline is the AMI web platform, which will help address broader challenges related to taxonomy, computer access, training, and international collaboration.

Already in use on three continents, AMI is being used across a wide range of ecology organizations. Through comprehensive insect monitoring, projects like this are giving entomologists the data they need to help inform policies related to land use, climate change, and conservation.

It is our hope that the data our methods provide to entomologists will help better inform land- use decisions and policy-making for climate-change adaptation and conservation.

David Rolnick, Assistant Professor, McGill University, Core Academic Member, Mila