A team of undergraduate students from Colgate University, New York, have developed an app called SealNet. This seal face-recognition software uses deep knowledge and a convolutional neural network (CNN) (artificial neural network used in image recognition) to distinguish one seal face from another. SealNet is invented to recognize the harbor seal, a species with a liking for posing on coasts in haulouts (living on the land). The team was led by Krista Ingram, a Biologist at Colgate University.
Image Credits: Pixabay
The team had to equip their software to identify the faces of the seals. But they have to recognize the mouth, the centre of the eyes and the nose manually. “I give it a photograph, it finds the face, [and] clips it to a standard size,” said Ingram.
For testing, the students collected more than 2,000 images of seals around Casco Bay, Maine, in two years. They checked the app using 406 distinct seals and discovered that its accuracy to identify seals’ faces is 85 percent. But the team enhanced SealNet’s database to include around 1,500 seal faces. “As the number of seals logged in the database goes up, so too should the accuracy of the identification,” said Ingram.
SealNet is not an error-free software. It recognizes seal faces in vegetation, other body parts and even rocks. Hence, the team advised it is best to check manually whether the seal face recognized by the app belongs to the real seal or not.
Other techniques such as tagging and aerial monitoring to chase seals can be highly expensive and invasive. But Ingram said that the SealNet app could be a non-invasive and useful tool for the researchers.
Ingram focuses on the accuracy of the SealNet app. The test conducted by the team showed that some harbour seals return to the same haulout places yearly. The other two seals, Petal and Clove nicknamed by the team, appeared at two different places together. “Increasing scientists’ understanding of how seals move around could strengthen arguments for protecting specific areas,” said Anders Galatius, an Ecologist at Aarhus University in Denmark.
“The software shows a lot of promise. If the identification rates are improved, it could be paired with another photo identification method that identifies seals by distinctive markings on their pledge,” said Galatius who is in charge of observing Denmark’s seal populations.
In the future, Ingram hopes to invent an app based on SealNet. The app could allow citizen scientists to participate in logging seal faces. This app could also be modified for other pinnipeds and cetaceans.
A detailed study has been published in the journal Ecology and Evolution.
Follow us on Medium!