Please view the main text area of the page by skipping the main menu.

New AI system speeds up identification of endangered coconut crabs

A person takes a smartphone photo of a coconut crab shell pattern in Motobu, Okinawa Prefecture, on Dec. 18, 2017. (Mainichi)

MOTOBU, Okinawa -- A research center here has introduced an artificial intelligence (AI) system capable of identifying endangered coconut crabs using their shell patterns, making the process far more efficient and precise than human work.

The Okinawa Churashima Foundation Research Center in Motobu, Okinawa Prefecture, has been surveying the coconut crabs in the foundation's Ocean Expo Park as part of its efforts to protect the crustaceans.

The coconut crab population is in decline globally due to overfishing and loss of habitat caused by development. According to Shinichiro Oka, a researcher at the center, the crustaceans are vulnerable to overfishing because they grow at a slow pace with an average life span of 50 years. They fall under the Japanese Ministry of the Environment's "Endangered 2 (Vulnerable Species)" category.

In 2006, the center began trying to identify an estimated 1,000 coconut crabs living in the park, and has since confirmed the existence of about 400 crabs there. But the process required patience and special know-how.

According to Oka, the shell patterns of coconut crabs differ from creature to creature, similar to the fingerprints of humans. In the past, researchers would take photos of any coconut crab they found, and try to identify those using previously taken photos. This method took at least 10 minutes to complete and was not free from mistakes.

The AI system is based on the outcome of joint research by Osaka University's Cybermedia Center and the Kyoto-based IT firm Cogito Inc. It utilizes deep-learning technology and studies the shell pattern characteristics of some 400 coconut crab shells using their photos.

When a newly taken photo of a crab is fed into the system, it can determine in an instant if that is a known individual. When a coconut crab is determined as previously unknown, the system registers its information. As the sample size gets bigger, the system's precision is expected to improve.

The research center is conducting a similar survey in adjacent waters for humpback whales, whose commercial fishing is prohibited. Humpback whales' tail fins have different shapes and patterns from whale to whale and thus can be used for identification.

The center has photographed around 1,500 tail fins, and cross-checked them with photographs of about 300 to 500 individuals observed each year. The process is time-consuming and it can sometimes take a full day to confirm just one fin, as it is difficult for the human eye to check the tail fin's jagged shape and pattern.

The research center is currently looking into cooperating with organizations such as Osaka University, and is trying to drastically reduce the checking time by applying the AI identification system used for coconut crabs.

For more information on the AI coconut crab identification system, visit: (Japanese language only).

(Japanese original by Masahiko Okamura, Planning Department)

Also in The Mainichi

The Mainichi on social media