The Shark Spotting project combines deep learning with marine science to facilitate research of sharks in coastal regions through drone footage.

A brief history

After initial exploration of building a model specifically with the purpose of recognizing sharks in the summer of 2019, two groups of students in Machine Learning classes in Fall 2019 worked independently to build this model. After learning from their trial and error, Professor Franz Kurfess applied to get our Shark Spotting with Drones project put into motion as a full-fledged research project with support from the Cal Poly SLO School of Engineering.

While these ideas were being developed, the shark lab at CSU Long Beach was spending hours rewatching and labeling drone footage in hopes of spotting a shark or two. Our project was initiated with the hope of creating a program that can recognize and categorize sharks, people, and other marine life from drone footage automatically, with Artificial Intelligence.

Where we are today

Our team is currently working on training the model to accurately identify sharks and people in drone footage. This information will allow scientists to increase knowledge on the relationship between people and the coastal ecosystems, including the relationship between humans and sharks. Applications to combine a neural network on a drone to identify sharks in real time to alert lifeguards is also being explored.

Programs We're Using

LabelBox

Label Box allows our team to easily label thousands of images in order to train a program for object detection.

PyTorch


Tensor Flow