Automating South Sea Pearl Phenotyping Through Masking and Deep Learning Algorithms

The commercial value for pearls come from their phenotypic traits such as colour, shape, size, lustre and surface complexion. The process of discerning these values currently involves highly skilled experts manually grading each successfully cultured pearl. This requires significant time, and in the context of genetic breeding program, relies on the availability of pearl graders for phenotyping. The aim of this research project is therefore to develop AI algorithms which automate the phenotyping process for genetic breeding, specifically on the size and shape of pearls.  The success of this project will enhance the breeding program, streamline grading processes, and can be potentially used in further authentication of pearl traits through block chains, fundamentally driving innovation in the Australian pearling industry.

Project Team and Collaborators:

JCU Team: Daniel McMahon, Carla Ewels, David Jones and Dean Jerry

Industry partners: Cygnet Bay Pearls