Internet of Birds has evolved to identify nearly 700 birds in India
January 2020
V&D: what is the thought process behind developing the Internet of birds app?
Sanjay Podder (SP):
India is a biodiversity hotspot and home to over 1,300 species of birds. Besides being beautiful creatures in their own right, birds are a ‘keystone’ genus and play a vital role as pollinators. They are important indicators of their environment and key to maintaining the overall ecological balance.
V&D: what has been Accenture’s contribution to the app’s design and content?
SP:
Accenture Labs, a key component of Accenture’s Innovation Architecture, has a team of applied R&D technologists who work to prototype and deliver ideas that drive strategic impact for both Accenture and its clients and business partners. Accenture Labs in Bangalore has provided pro-bono services to BNHS to design and build the Internet of Birds platform.
This has been done as part of Accenture’s Tech4Good initiative, under which Accenture uses emerging digital technologies and works with leading social Innovators, academia, startups and government to address sustainable development goals across education, health, environment, inclusion and diversity.
V&D: what are the technologies used to power the app?
SP:
The Internet of Birds is a cloud-based platform that leverages AI, machine learning and computer vision. It runs on an Accenture-developed image recognition and deep learning platform to swiftly and accurately identify bird species from digital photos that are uploaded by users.
The platform has been trained on birds found in the Indian subcontinent using a convolutional neural network, a deep learning algorithm that can take in image and assign importance to various aspects in the image, with the ability to differentiate between images. A unique feature is that it uses AI at the edge to allow bird watchers use mobile phones to identify birds, even in remote locations such as deep jungles with no internet connectivity.
The platform also uses a unique citizen crowd sourcing approach under which bird watchers can contribute to the platform by uploading information on rare birds that they come across. Data from BNHS helps confirm the identity of the birds and trains the application’s deep learning model. Each time a picture is contributed to the system, it teaches itself, increasing the platform’s accuracy in the recognition of bird species.