Citizen science meets AI

Your next social media post could become a data point in an animal study. Recent scientific naturalist studies are using artificial intelligence to sift through thousands of photos and extract data about animal sightings.

One research group is looking for photos of Anopheles stephensi, a malaria-carrying species of mosquito adapted to urban settings. It has spread to more than six African countries in the last 10 years. A group of researchers from the University of South Florida and the University of Antananarivo in Madagascar created an AI tool that can detect Anopheles stephensi in photos. They took thousands of smartphone photos of mosquitoes and trained an AI model to identify the species using a photo recognition algorithm.

Now the tool can automatically analyze photos of mosquitoes submitted to citizen science apps like iNaturalist or NASA’s GLOBE Observer. Recently, it identified Anopheles stephensi in a photo from Madagascar taken five years ago, before the species was believed to be present in the island country.

“Traditional mosquito surveillance such as trapping is expensive, requires expertise, and time,” said Ryan Carney, the project’s lead researcher. By contrast, the AI tool facilitates a citizen science approach to disease prevention, which should be a “priority for both local communities and international health bodies like the World Health Organization.”

Carney said getting the word out can be a problem: “For these tools, one of the limitations is that people are simply unaware that the citizen science apps exist, or that photos of mosquitoes are needed by scientists.” To encourage participation, Carney’s team hosts workshops in elementary schools and universities in Madagascar.

A group of scientists studying sharks has found a different solution to a similar problem. People may not know that scientists need shark photos, but they’re usually excited to post their shark sightings on social media. Marine biologists from Virginia Tech and Stanford University have created an AI tool that doesn’t rely on photos submitted to science-related applications. Instead, it uses Instagram posts, Facebook posts, and other photos posted online.

The AI platform, called “sharkPulse,” relies on web data mining, the process of computers automatically collecting and organizing information from thousands of sources across the internet. In the case of sharkPulse, the AI tool collects photos of sharks in addition to time stamps, location data, and species information. A group of marine biologists validates each photo before adding it to a searchable database.

“With cameras in nearly everyone’s hands, our encounters with the ocean are being recorded more than ever,” said Jeremy Jenrette, a Ph.D. candidate working on the project. “SharkPulse turns scattered signals into knowledge.”

SharkPulse currently has over 100,000 records featuring over 285 shark species. The tool also can analyze historic photos, which provide data points as far back as the 1890s. Using the records, marine biologists have identified new hot spots of shark activity, notably white shark hot spots in the Mediterranean. The team is currently improving sharkPulse’s multilingual data mining capacities and working to integrate other data sources like fishery records.

SharkPulse could be adapted to create similar platforms for other species. “This is about creating an always-on pulse monitor for the ocean,” lead researcher Francesco Ferretti explained. “The more we see, the more we can do to protect.”