Published on 15 May 2022

Identifying birds with the aid of artificial intelligence

Two Nanyang Junior College students created an online identification application that names birds based on uploaded images.

Source: The Straits Times © SPH Media Limited. Permission required for reproduction

Identifying birds with the aid of artificial intelligence

Two Nanyang Junior College students created an online identification application that names birds based on uploaded images. REHMAT KAUR THAKRAL reports.

Unless they are bird enthusiasts, few people would be able to name the various species of birds they come across.
Two students from Nanyang Junior College have devised a digital solution to this problem.

Edwin Ong and Liow Shang Yu, both 17 and in their second year, came up with an innovative Web application that uses artificial intelligence (AI) to identify different birds.

To identify a bird, a user has to upload an image of it. The application will then suggest the likely species of the bird, the probable accuracy of the prediction, and a sample picture of what the app’s AI model thinks the bird looks like.

The user will then be able to learn more about the bird by choosing an option that redirects to the Singapore Birds Project website, which provides details such as a bird’s common behaviour.

The boys worked on this application from April last year to January as part of the Nanyang Research Pro- gramme, an annual outreach pro- gramme by Nanyang Technological University for students in their first year of junior college.

“We were hoping that we would be able to benefit bird watchers and, maybe, contribute more to conservation efforts because birdwatchers play a very vital role in conservation efforts,” said Shang Yu. Birdwatchers provide huge amounts of data to conservation researchers.

During the project, the students were helped and guided by Dr Kenneth YT Lim, 55, and his team.

“Over the past year, I’ve been very impressed with how proactive the pair was in looking after their own learning,” said Dr Lim, a senior re- search scientist at the National Institute of Education.

The boys faced many challenges throughout the process as they had little prior knowledge of artificial intelligence and Web programming.

“The maths was university level, and, with our busy schedules, it was difficult to allocate time for us to learn the fundamentals of artificial intelligence,” said Edwin.

Regardless, they persisted. They tried to solve the issues independently and referred to online forums and YouTube tutorials for solutions.
From April last year to December, the boys used deep learning to develop the identification model of the application, which identifies the species of the birds from uploaded pictures.

Deep learning is a method in artificial intelligence that teaches computers to process data in a way that imitates how humans gain knowledge.
Using datasets from Google Im- ages, the duo uploaded an average of 65 images per bird for the algorithm to learn how to recognise the birds. They uploaded profiles of 41 birds to the identification model.
The app does have some limitations, though. For example, as they had a limited time to work on the app, it currently can identify only the birds they added to the model so far, but there are over 300 birds native to Singapore.
Furthermore, if an uploaded im- age is too grainy or too low in resolution, the application’s identification model may not be able to clearly pick the bird out from the background.
“Our app isn’t the end product,” Edwin added. “It’s an idea that can be improved on to help serve its purpose for conservation.”
The duo‘s next step for the application, which is not yet available to the public, would be to expand its database, provided they have the time and resources. Once they have made the necessary improvements, they hope to release it.