Download FAR 2019/58 Feasibility of automating otolith ageing using CT scanning and machine learning

This study explored using CT scanning and machine learning to automate fish ageing. CT scanning resolved banding patterns on the surface of snapper, hoki and ling otoliths, but resolution and contrast were too low to detect outer bands in images through otolith cores. A convolutional neural network was used to estimate snapper and hoki ages from otolith images, and after limited training obtained promising consistency with human ageing. Further development of both techniques is recommended.

AUTHORS: Moore, B.R.; Maclaren, J.; Peat, C.; Anjomrouz, M.; Horn, P.L.; Hoyle, S.