La Jornada: They propose to combat marine wildlife trafficking with AI

Madrid. Many commonly trafficked marine wildlife items, such as shark fins, can be hidden in luggage or packages and transported across borders relatively easily, undetected, so to address this, scientists at Macquarie University, Australia, used artificial intelligence to develop an algorithm capable of detecting samples of commonly trafficked marine creatures (shark fins, seahorses and sea cucumbers) with an accuracy of 92 percent.

“The wildlife trade is cruel and unethical,” says Dr Vanessa Pirotta of Macquarie University, lead author of the new paper published in Frontiers in Ocean Sustainability. “For many, this may be the first time they have heard of illegal trafficking in marine wildlife, something that is not mentioned as much as rhino horn or elephant ivory. We take advantage of this World Oceans Day (celebrated yesterday) to make this problem visible.”

It is estimated that the illegal trade in marine fauna moves billions of dollars each year and represents a serious threat to endangered animals. The transport of animals for human consumption, medicinal, ornamental or as pets endangers the survival of populations that live in a precarious balance, while animals that are trafficked alive could escape and become invasive species in other ecosystems.

However, detecting this type of smuggling is not easy, which makes it difficult not only to stop it, but also to quantify its environmental impact.

The team repurposed existing X-ray CT scanners, which are used at many airports to detect explosives or biosecurity threats. These scanners take multiple X-rays of a single object, creating a three-dimensional image of its contents.

Shark fins and seahorses

By using a neural network to train an algorithm capable of recognizing commonly trafficked species in these images, the scientists hoped to create a system that would automatically mark suitcases for inspection.

The researchers chose to work with shark fins, seahorses and sea cucumbers. Shark fins are in high demand as food, while dried seahorses are marketed for traditional medicine.

Smuggling of sea cucumbers is less frequently recorded, although they are often illegally overfished, so it is believed that their smuggling is more common than can now be demonstrated.

The team performed 298 scans of 20 sea cucumber, 30 seahorse and 18 shark fin samples, many of them from wildlife trafficking seizures. Five scans were created for each sample in different positions and contexts, as well as scans containing multiple samples.

The scientists also scanned samples in conditions that mimic smugglers’ tactics (wrapped in cans or clothing or hidden in children’s toys) and added some of their scans to CT images of suitcases that had been scanned without contraband merchandise, a technique called threat imaging.

These images were used to train the algorithm to recognize shark fins, sea cucumbers, and seahorses, and then tested the algorithm on a subset of shapes that had never been provided to it before.

The algorithm had an overall effectiveness of 92 percent: 95 percentage points in detecting shark fins, 96 in detecting seahorses and 86 percent in detecting sea cucumbers. The false positive rate was 13 points: 2 percent for shark fins, one for sea cucumbers and 9 percent for seahorses.