AI-Driven Accelerometer-Based Bird Activity Recognition
TimeTuesday, July 12th6pm - 7pm PDT
LocationLevel 2 Lobby
Event Type
Networking Reception
Work-in-Progress Poster
DescriptionTo better understand species like birds, researchers need to gain insight into their activities and behaviors. From the domain of humans, we know that accelerometer-based activity recognition is a well-working approach. With this work, we take the so-far human-centric approach and explore the use of accelerometer data to classify bird activities. We build different neural networks (CNN+DNN and CNN+RNN), that are capable of successfully distinguishing four activities of birds (swimming, flying, on ground, and foraging) with an accuracy of so-far 70-80%.