This app offers to save endangered sea turtles

Named SEE Shell, the application uses machine learning (or machine learning) to analyze a photograph of an object with a turtle shell pattern to determine with 94% accuracy whether it is real or false. It is the first mobile application to use computer vision to combat the illegal wildlife trade, according to Alexander Robillard, a doctoral student at the Smithsonian’s Data Science Laboratory who created the computer model that drives the application. . It will help inform buyers with conservation mind as well as law enforcement. (The app can be downloaded from the Apple App Store and Google Play.)

According to Nahill, turtle shell sales on the black market persist in at least forty countries, mostly in Central America and Southeast Asia. The vast majority of these illicit sales are due to tourists buying turtle trinkets at gift shops and souvenir stalls.

The application is also very useful for the large amount of information it can provide access to. All images are uploaded to a private, centralized database, along with their GPS coordinates, allowing SEE Turtles to identify where the major illicit sales locations are located.

“If we encourage a few hundred travelers to actively use it, collect data, and avoid buying a real turtle shell, that’s good enough,” Nahill says. SEE Turtles plans to promote the free app through social media campaigns and collaborations with other conservation organizations.

Marine ecologist Emily Miller, who did not participate in the development of the app but was the author of a 2019 article on the global scale of the hake trade, explains that although many groups of around the world collect data on the trade in these animals, “one of The main obstacle that makes it difficult to answer research questions is the step of consolidating, formatting and organizing all the data.” Having a larger centralized database “will be incredibly useful in understanding global business patterns,” he adds.


Robillard worked with Nahill’s team to collect 4,000 images of real and fake turtle products. The researcher introduced these images into his computer model, which analyzed the pixels of each of them in order to determine the differences in shape and color between a real scale and a false one.

According to Nahill, one of the main differences is that the actual turtle patterns are random. Counterfeit products, on the other hand, usually have stains with uniform edges or the same pattern on different items sold together. The orange tone of the false scale also tends to be uniformly translucent.

Nahill and Robillard know how to distinguish good from evil, but without the application it may take years of practice for a non-expert to develop this skill. “I like to tell people that they have Brad [Nahill] in your pocket! Robillard says of the app. Machine learning and computer vision “can do any visual task that a human can do, but more efficiently and quickly,” he adds. (For my part, I tried SEE Shell with two pairs of turtle print glasses and immediately identified them as fakes).


Thanks to the app, scientists have already discovered turtle shell products that they never knew existed: for example, cocktail mixing sticks or even cockroaches for cockfights.

The app will be very useful for local conservation groups. Prior to its launch, the Tortugas del Mar Foundation, a turtle protection group in Cartagena, Colombia, had already successfully pushed local law enforcement to crack down on trade, reducing it by almost 80% in the region. But according to Nahill, authorities only patrol if a member of the group accompanies them to help them identify illegal products. Tortugas del Mar plans to train law enforcement to use the app, which will allow them to work more quickly and independently.

David Godfrey, executive director of Sea Turtle Conservancy, which works specifically to protect hawksbill turtles in Panama, a major trading post, believes that if tourists use SEE Shell, it will be tantamount to equipping “an army of conservationists. “. difficult for people to sell [l’écaille de tortue] on the black market. Now that people can immediately identify the actual scale, sellers can think twice before offering it, he adds.

With the help of the World Wildlife Fund (WWF), SEE Turtles also aims to apply this technology to online platforms, which have seen an increase in the illegal wildlife trade in recent years. Among other things, Facebook and eBay filter out illegal products by detecting ads that include banned keywords, but filters can be easily circumvented. “When we know, no one has implemented any kind of visual system,” says Nahill.

This visual learning technology could be adapted to other wildlife materials, for example, to distinguish real bones from fake ones. According to Robillard, the ability to identify real elephant ivory in an instant would be especially useful, but it would be more complicated than for the turtle: a key indicator is to look at the internal cross lines to determine if the ivory is genuine or not. no, which is not possible with a simple photograph.

However, according to him, “there is a whole range of possibilities to apply machine learning to conservation issues.”

Leave a Comment