Have you ever appeared at a seal and assumed, Is that the identical seal I saw yesterday? Very well, there could shortly be an app for that centered on new seal facial recognition technology. Recognized as SealNet, this seal facial area-discovering method was developed by a workforce of undergraduate college students from Colgate College in New York.
Taking inspiration from other technology tailored for recognizing primates and bears, Krista Ingram, a biologist at Colgate University, led the pupils in creating software that employs deep discovering and a convolutional neural network to convey to just one seal face from yet another. SealNet is customized to establish the harbor seal, a species with a penchant for posing on coasts in haulouts.
The team had to train their software to recognize seal faces. “I give it a photograph, it finds the deal with, [and] clips it to a regular sizing,” says Ingram. But then she and her pupils would manually discover the nose, the mouth, and the centre of the eyes.
For the undertaking, workforce associates snapped additional than 2,000 photographs of seals close to Casco Bay, Maine, in the course of a two-calendar year period of time. They examined the software program making use of 406 diverse seals and discovered that SealNet could effectively detect the seals’ faces 85 percent of the time. The crew has considering the fact that expanded its database to incorporate close to 1,500 seal faces. As the selection of seals logged in the databases goes up, so far too need to the accuracy of the identification, Ingram states.
As with all tech, even so, SealNet is not infallible. The software observed seal faces in other body areas, vegetation, and even rocks. In a person case, Ingram and her college students did a double choose at the uncanny resemblance concerning a rock and a seal confront. “[The rock] did appear like a seal face,” Ingram suggests. “The darker sections were about the exact same distance as the eyes … so you can have an understanding of why the application observed a experience.” Consequently, she says it is generally greatest to manually test that seal faces identified by the program belong to a actual seal.
Like a weary seal hauling itself onto a seaside for an involuntary photograph shoot, the question of why this is all important raises by itself. Ingram believes SealNet could be a handy, noninvasive software for scientists.
Of the world’s pinnipeds—a group that features seals, walruses, and sea lions—harbor seals are thought of the most widely dispersed. However knowledge gaps do exist. Other methods to observe seals, this sort of as tagging and aerial checking, have their constraints and can be hugely invasive or expensive.
Ingram factors to internet site fidelity as an aspect of seal habits that SealNet could get rid of extra gentle on. The team’s trials indicated that some harbor seals return to the exact haulout web pages yr soon after 12 months. Other seals, having said that, these as two animals the team nicknamed Clove and Petal, appeared at two distinctive web-sites alongside one another. Raising scientists’ knowledge of how seals shift around could fortify arguments for safeguarding unique parts, states Anders Galatius, an ecologist at Aarhus University in Denmark who was not included in the job.
Galatius, who is responsible for monitoring Denmark’s seal populations, claims the application “shows a great deal of promise.” If the identification prices are enhanced, it could be paired with another image identification method that identifies seals by distinct markings on their pelage, he claims.
In the long run, immediately after even more tests, Ingram hopes to build an application based mostly on SealNet. The application, she states, could quite possibly let citizen researchers to contribute to logging seal faces. The plan could also be adapted for other pinnipeds and possibly even for cetaceans.
This write-up is from Hakai Magazine, an on line publication about science and modern society in coastal ecosystems. Study much more tales like this at hakaimagazine.com.
Similar stories from Hakai Magazine:
Suggested Video clips