Urbanization is one of the key drivers of global biodiversity decline, with insect species populationspredicted to be one of the ecosystem components hit hardest. In this research, MIT and AMS Institute researchers propose a novel computer vision model and method to quantify insect population dynamicsin real-time to promote urban biodiversity.
Amsterdam aims to “rigorously green” the city—implementing new green infrastructures, fosteringclimate-adaptive solutions, and seeking to bolster biodiversity. However, the city has little data showinghow these interventions, air pollution, and climate change may affect biodiversity. So, how can wemeasure biodiversity in a city? And what constitutes urban biodiversity?
One crucial measurement hereis the health and abundance of insect species. Insect biodiversity and abundance are in global decline, potentially leading to a crisis with profoundecological and economic consequences. The problem's urgency is enormous, yet we need to learn more about the role of cities in this crisis.
In new research published in Nature Scientific Reports, Towards real-time monitoring of insect species populations, researchers working at the MIT Senseable City Amsterdam, a collaboration between MIT Senseable City Laboratory and AMS Institute, propose a computer vision model that works towards "multi-objective insect species identification" in real-time and on a large scale. It leverages an image data source with 16 million impressions and allows a quick and open-access method to develop visual AI models to monitor insect species across climatic regions. This model will first be deployed in Amsterdam, but eventually, the technology could be used to create data-driven insights for other cities seeking to safeguard or promote biodiversity.
The AI model used in the study focuses on insect species in the Western European region. This model, part of the B++ project, trained on 1.54 million web-scraped images, can classify 2,584 insect species and could be deployed on images collected from high-definition cameras in urban, suburban, agricultural, and natural areas. For scalability to other geographic regions, the researchers can present a code repository that uses an existing 16 million image dataset to train custom AI models for local insect species of interest.