Train artificial intelligence to identify fish in Kakadu National Park (BRUVNet)

The Supervising Scientist in the Northern Territory, Australia monitor freshwater fish communities in Kakadu National Park using underwater video cameras, collecting 100’s of hours of fish videography every year!

This represents a significant volume of video that trained ecologists must manually process to characterise fish community composition.
We can automate this process by training artificial intelligence models to identify and count the fish for us.

This however requires labelled images of fish with lines drawn around them to tell the model where fish are in the image and what species it is.
Our scientists have partnered with machine learning engineers from Microsoft to develop this technology.

This is where we are calling on citizen scientists of the world to help us annotate thousands of fish images. It will enable us to build a powerful model used to count fish with high accuracy.

This dataset will be made public and enable other fisheries scientists around the country to automate counting of their own fish!

If you are interested in contributing by drawing polygons around our Kakadu fish species, contact the email address or go to www.bruvnet.org to get set up with your unique set of fish images and start labelling!

All you need is a computer with a web browser.