Challenge:
Determining the sex of sturgeon requires invasive, costly manual inspection. Aquaculture producers need a faster, cheaper method to make selective breeding and stock management decisions.
Approach:
Developing a computer vision system trained on tens of thousands of high-resolution images of male and female sturgeon. The model is designed for deployment on NVIDIA Jetson edge hardware at aquaculture facilities, with an over-the-air update pipeline so accuracy improves continuously as new data is collected.
Result:
Active three-year project (2025-2028) funded by the Western Regional Aquaculture Center (WRAC) at $465,851. Lead PI: Edwin Solares. Institutions: UC Davis and University of Washington.
Expected impact: Reduce per-fish phenotyping cost and processing time versus manual inspection, enabling aquaculture producers to make faster selective breeding decisions at scale.
