Download FAR 2021/69 Development of deep learning approaches for automating age estimation of hoki and snapper

This study used deep learning to provide an automatic estimation of age for hoki and snapper through a convolutional neural network (CNN). A reference library of otolith images from ~1060 hoki and 520 snapper was generated for use in the CNN. Results from models using these images suggest that deep learning has the potential to support the automation of fish ageing, although further research is required to build an operational tool useful for routine fish ageing.