Covering Disruptive Technology Powering Business in The Digital Age

image
WWF Indonesia Using Facial Recognition on Orangutans
image
September 26, 2019 News

 

While facial recognition technology has permeated various industries and sectors for different use cases around the world, there have been many concerns about how effective this tool can be. While facial recognition biometrics has increased efficiency for identifying individuals, many feel the technology is still far from perfection. From immigration checkpoints to unlocking mobile devices to analysing emotions, facial recognition technology is used almost everywhere today.

But little did we ever expect the same technology to be used on animals as well. WWF Indonesia and AWS are currently on a pilot stage of using facial recognition technology to study the behaviour and habitat movements of orangutans in Punggualas, Kalimantan, Indonesia.

According to Edwin H Chaidir, IT Manager, WWF Indonesia, they have been working on a machine learning approach through facial recognition on the orangutans.

“Unlike birds and turtles that can be tagged, orangutans are different. If we tag them, they will remove the tags. The orangutan population is declining rapidly in Indonesia. Every year we lose about 2000 orangutans to deforestation and poaching. Today we are merging science and technology from the ground to the cloud.”

Using AWS, researchers would upload images onto the data set on the S3 browser and have it processed on the S3 bucket. An Amazon S3 bucket is a public cloud storage resource available in AWS Simple Storage Service (S3), an object storage offering. Amazon S3 buckets, which is similar to file folders, store objects which consist of data and its descriptive metadata.

From there, researchers use Amazon SageMaker to generate models. AWS SageMaker labels the orangutan by creating a box around the face. It needs about 30 images to recognise each orangutan. The accuracy rate ranges from 35% to 95%. The more data they have, the higher the accuracy of facial recognition of the orangutan.

The models are then deployed via AWS Inference to AWS SageMaker Endpoint onto the Amazon API Gateway. WWF also uses AWS Lambda functions with the data from the Amazon API Gateway and is able to come up with the desired results. The entire process is still in its pilot stages, but WWF Indonesia believes they are heading in the right direction.

In the future, WWF hopes to use camera traps to capture more images of the orangutan. If successful, they believe the same method can be used on other animal species like tigers and rhinos as well as whale sharks.

WWF Indonesia has been working with AWS since March this year, and the input and effort from AWS to help in this pilot project have been successful. AWS offers a variety of services as they have the algorithms, making it easier for WWF.

“We have a lot of data, but we are not sure what to do with it. AWS tech has helped with this. By keeping track of the metadata, hopefully, we can understand the orangutans better. Technology has revolutionised the way we manage our environment and protect biodiversity.”

(0)(0)

Archive