Scaling Up Ecological Monitoring with AI: How Supercomputing Is Unlocking the Value of Autonomous Acoustic Sensing

August 23, 2024

1:00 pm – 2:00 pm Eastern time

Speaker: Sam Lapp, University of Pittsburgh

Automated acoustic recording devices monitor wildlife at an unprecedented scale. Using hundreds of sensors, bioacoustic monitoring transforms our understanding of birds, bats, and frogs, aiding biodiversity management. Analyzing vast data (e.g., 70 Tb = 100,000 hrs of audio) is challenging. The Kitzes Lab at the University of Pittsburgh uses Bridges2 supercomputing to overcome this. Supercomputing nodes enable AI to provide new biological insights from recordings. This seminar highlights three projects: using CNNs on GPU nodes to reveal habitat needs for Great-horned Owls; extreme-memory nodes detecting the rare Panamanian golden frog; and CPU/GPU parallelization tracking Pennsylvania songbirds’ territory use. These projects show how Pittsburgh Supercomputing Cluster enhances AI in bioacoustic data analysis, improving biodiversity understanding and conservation.

About the speaker: Sam is a machine learning and sound geek with a passion for biodiversity conservation. He asks: How can we prevent biodiversity loss? Where do we develop reciprocity between humans, technology, and ecosystems? His current research focuses on developing and applying AI methods in bioacoustics, the strategy of studying ecology through the sounds produced by living things. His projects include developing the open-source Python package OpenSoundscape and applying machine learning methods to the conservation of birds, frogs, and insects. Sam studied engineering, music, and sound design at Penn State and is currently working on his Ph.D. in the Kitzes Lab at the University of Pittsburgh.