The app, called “SEE Shell,” used machine learning to determine whether a tortoiseshell item was 94% real or fake. The app looks for distinct patterns that show if a substance is natural. According to Alexander Robillard, a Smithsonian OCIO Data Science Lab employee who developed the computer software behind the app, the technology is the first mobile app to use computer vision or “machine vision” to combat the illegal trade in wildlife products and can be used by environmentally conscious buyers and law enforcement officers over the years. Both. (You can download the app from the Apple App Store and Google Play).
According to Nehill, there is an active market for turtle shells in at least 40 countries. The lion’s share of these markets is in Central America and Southeast Asia, and the vast majority of illegal turtle shell goods are sold in boutiques and gift shops to tourists on vacation in these areas.
The app is also valuable due to the great wealth of information it can provide. Each captured image of a turtle shell object will end up being uploaded to a central and private database, including the image’s GPS data, so that SEE Turtles employees can see exactly the hot spots of this illegal trade.
“Even if there are only a few hundred travelers actively using the app, collecting data and not buying these products, that’s a huge plus,” says Nahil. SEE Turtles plans to distribute the app for free using social media campaigns and collaborations with environmental organizations.
According to Emily Miller, a marine ecologist who was not involved in developing the app but wrote a scientific paper in 2019 on the global trade in hawksbill shells, there are now multiple environmental groups in the world that collect data on the trade. In the sea turtle shells, collect the hawksbill turtle. “One of the biggest obstacles to finding answers to research questions is to standardize, coordinate, and organize all this information,” she says. Building a larger, centralized database “would be an incredible contribution to our understanding of the global trade routes of hawksbill shells,” Miller said.
Learn to save turtles
The Robilard and Nahill team collected 4,000 images of natural or synthetic turtle shell products. He then inserted the images into a computer model, which analyzed the pixels in each image and taught himself to distinguish the differences in colors and shapes between real and fake products.
An important difference, according to Nehill, is the fact that the patterns in real turtle shells are completely random. Resin products often have edges with repeating patterns or patterns that can be found on many items from lots for sale. Also, the orange color seen in many fakes often has a consistent transparency.
Nahil and Robyar are now very good at distinguishing between real and fictitious tortoiseshells, but without applying them, it might take an inexperienced person years to master this skill. I always say it’s Brad [Nahill] Easily beat with this app! Robilard says. He explains that machine learning and computer vision “can perform a visual task that humans can do, but more efficiently and faster.” (I tested SEE Shell on two tortoiseshell-framed glass—and immediately found that they were both synthetics.)
Rods and traces
Using the app, scientists have already discovered tortoiseshell products they didn’t even suspect existed, such as cocktail stirrers and cockfighting spores.
The application will also be of value to environmental groups. Even before the new technology came online, Fundación Tortugas del Mar, a sea turtle conservation group in Cartagena, Colombia, had achieved success by providing local law enforcement officials with the app, allowing trade in turtle shell products in the area to increase by nearly eighty percent. But according to Nehil, authorities will only send a patrol if they are accompanied by someone from the environmental group to help identify illegal products. Tortugas del Mar plans to train law enforcement officers on how to use the app so they can act more quickly and independently of the group.
David Godfrey, director of the Sea Turtle Conservancy, a group that works, among other things, to protect loggerhead turtles in Panama (a turtle shell trading hotspot), the SEE Shell can also be used by tourists. In this case, “it would be as if an entire army of environmentalists were hunting down sellers of these illegal products.” If people can instantly distinguish between a natural product and synthetic versions, sellers will likely think twice before selling these things, he says.
With the help of the World Wildlife Fund for Nature, SEE Turtles wants to introduce the new technology also on online platforms where a significant increase in the supply of illegal wildlife products has been observed in recent years. Facebook, eBay, and other companies try to block this offer by filtering ads for suspicious keywords, but it’s easy to bypass these filters. “As far as we can tell, no one has done anything visually yet,” Nahil says.
“Machine vision” technology can also be applied in principle to distinguish between other toy products and counterfeit products. For example, the ability to identify real elephant ivory would be especially valuable, according to Robillard, although it would be more difficult than identifying a true tortoise shell, because one of the main features of real ivory is its internal hatching that is unrecognizable in photographs. However, he says, “there is a world of possibilities for applying machine learning in the environmental field.”
The National Geographic Society is dedicated to highlighting and protecting the wonders of our world, and has historically funded the work of National Geographic researcher Brad Nahill to protect sea turtles. Learn about the Society’s support for National Geographic researchers who describe and protect important species.
It also supports the National Geographic Society Wildlife Watch, our investigative journalism project that draws attention to wildlife crime and wildlife exploitation. Send tips, feedback, and story ideas to NGP.WildlifeWatch@natgeo.com.
This article was originally published in English at nationalgeographic.com