Project Overview (& PhD Embargo)

Behavioural changes are commonly first indicators of disease in fish, and can be used to detect disease status before more overt symptoms (e.g. lesions, changes in colouration, other clinical symptoms) become apparent. This project aimed to determine whether computer vision models could be trained to detect disease status in aquaculture-relevant fish from behaviour and generalised indicators of disease like melanonis or changes in body positioning.

The video footage used in this project was collected as part of a vaccine trial in rainbow trout conducted at the University of Queensland. The model was trained on top-down videos of industrial aquaculture tanks and individual-level fish annotations. Swimming trajectory and cropped images of each fish were used as inputs to the model, and the model was trained to predict disease status (affected vs unaffected) of each fish.

This research project forms the foundation of one of my three experimental chapters in my PhD thesis. As my thesis is currently under review, and I am awaiting my viva, I cannot publish research results here at present as this is not an industry-linked work. However, I will update this project page with more details once my doctorate is conferred and a manuscript of this experiment is submitted. In the meantime, if you are interested in learning more about this project, please don't hesitate to reach out to me directly .