Congratulations to Sammamish senior Manoj Sarathy, who was awarded Most Original and Best in Category: Science, Technology, and Society from MIT’s Inspire competition for his project on using computer vision to aid conservation biologists!
Sarathy’s award-winning project was a machine learning model that he designed, which helps conservation biologists get their data more quickly than they have historically.
During the last 10 years, conservation biologists have increasingly relied on remote camera traps to study wildlife presence and behavior. These remote camera traps are digital cameras enclosed in a weatherproof case triggered by either heat or motion.
Sarathy’s work on his project has extended over a couple of years.
Last year Sarathy created a school-based conservation team at Sammamish, which has since deployed two such remote camera traps in the Cascades on behalf of Conservation Northwest. Conservation biologists at this organization use data from these images to determine where to apply their limited rewilding resources, such as to reintroduce animals or build overpasses or underpasses.
Based on the club’s experience and discussion with conservation organizations, Sarathy learned that 75 percent or more of the images camera traps capture are false positive, meaning that no animals appear in the image, which is because most camera traps detect motion and foliage or trees that move in the wind can trigger the remote camera traps.
To further research and develop, Sarathy reached out to a carnivore researcher at the Woodland Park Zoo, who supplied him with tens of thousands of images to help develop the machine learning model. He then used a computer to create the model, which now achieves more than 90 percent accuracy in separating false positive images from other images containing animals or humans. Several conservation organizations are now working with Sarathy, including in the Bay Area and internationally.
MIT is awarding Sarathy $2500 for his achievement. Additionally, he presented his paper and findings at the Wildlife Society’s 25th annual conference Oct. 7-11 in Cleveland, Ohio.
To learn more, visit the competition’s website.