This repository contains code to compare three segmentation pipelines applied to ISIIS images, against a ground-truth segmentation generated by a human operator.
https://github.com/ThelmaPana/segmentation_benchmark
The results of this comparison are published in Frontiers in Marine Science: 10.3389/fmars.2022.870005. The war machine we have developed – based on a CNN detector – generates the best results, but it processes data with the slowness of a tank.