Using OWL Integration's Cluster Duck Protocol paired with their DuckLink IoT devices we use machine learning at the edge to monitor for sharks. Using a USB camera connected to a Google Coral Board we take images and classify those images into two classes, shark or non-shark. The images taken by the camera are run through a Convolutional Neural Network (CNN) that was trained with over 5,000 images of different species of shark, dolphins, sea turtles, and fish. The output of the CNN is the class of the image along with the confidence level of the model. This output is then sent from the Coral Board to a DuckLink device and will then be sent off to a database for other applications to use or just to keep as a record.

Our team collaborated with Electrical and Computer Engineering 16 on this project.