LookingForSeagrass with Semantic Segmentation

This is the code which corresponds to the Seagrass Semantic Segmentation paper which was puplished at the IEEE/OCENs conference 2019 in Marseille

Documentation for usage

Tensorflow is structured in it`s components:

Recreated network architectures from papers descriptions:

Usage:

import main
main.deepSS(<MODE>, <Neural Network Name>)

Whereas MODE(first parameter) can be either train, predict, eval or sanityCheck. The names of the neural networks are the filename without the .py extension. Information about the modes:

Example call:

import main
main.deepSS("eval","deeplabV3plusSS")

The model is saved into a folder named models which is two directories above relative to the main file. The same for the tensorflow log files which are located in ../../logs and the dataset folder in ../../data.

Main JSON config file:

Dataset JSON config file: