Base copy of the article
Have you ever looked at a seal and thought: Is this the same seal I saw yesterday? Well, there may soon be an application for this, based on a new technology for recognizing seal faces. Known as SealNet, this seal face detection system was developed by a team of students at Colgate University in New York.
Inspired by another technology adapted for recognizing primates and bears, Christa Ingram, a biologist at Colgate University, guides students in developing software that uses deep learning and a convolutional neural network to distinguish one seal’s face from another. SealNet is adapted to identify the harbor seal, a species that tends to pose along the coast in howls.
The team had to train their software to identify seal faces. “I give him a picture, he finds the face, [and] it cuts it to a standard size, ”says Ingram. But then she and her students will manually identify the nose, mouth and center of the eyes.
For the project, team members took more than 2,000 photos of seals around Casco Bay, Maine, over a two-year period. They tested the software using 406 different seals and found that SealNet could correctly identify seal faces 85 percent of the time. Since then, the team has expanded its database to include about 1,500 seals. As the number of stamps registered in the database increases, so does the accuracy of the identification, Ingram said.
SealNet developers have trained a neural network to distinguish port seals using images of 406 different seals. The photo was provided by Birenbaum et al.
However, as with all technologies, SealNet is not infallible. The software saw the faces of seals in other parts of the body, vegetation and even rocks. In one case, Ingram and her students took a double look at the incredible resemblance between a rock and a seal’s face. “[The rock] it looked like the face of a seal, ”says Ingram. “The darker parts were about the same distance as her eyes, so you can find out why the software found a face.” seal.
Like a tired seal dragging itself to the beach for an involuntary photo shoot, the question arises as to why all this is necessary. Ingram believes that SealNet can be a useful, non-invasive tool for researchers.
Of the world’s pinnipeds – a group that includes seals, walruses and sea lions – harbor seals are considered the most common. And yet there are gaps in knowledge. Other seal tracking techniques, such as marking and aerial surveillance, have their limitations and can be highly invasive or expensive.
Ingram points to the site’s loyalty as an aspect of seal behavior on which SealNet can shed more light. The team’s experiments showed that some port seals return to the same storage places year after year. However, other seals, such as two animals, the team nicknamed Carnation and Petal, appeared in two different places together. Increasing scientists’ understanding of how seals move could strengthen arguments for protecting specific areas, said Anders Galatius, an ecologist at the University of Aarhus in Denmark who was not involved in the project.
Galatius, who is in charge of monitoring Danish seal populations, says the software “shows a lot of promise.” If the level of identification improves, it could be combined with another method of identifying photos, which identifies seals with distinctive markings on their skin, he said.
In the future, after further testing, Ingram hopes to develop an application based on SealNet. The app, she says, could allow civilian scientists to contribute to the registration of seal faces. The program can be adapted for other pinnipeds and probably even for cetaceans.
Add Comment