A Posidonia oceanica Semantic Segmentation model implemented in tensorflow.
Folder organization:
- preprocess: contains scripts to preprocess the images and ground thruts, resize, change extension, change color,...
- vgg16fcn8: contains all the original network files, forked from https://github.com/MarvinTeichmann/KittiSeg.
- evaluation: contains scripts to binarize the output of the network, evaluate its performance and view the missclasified areas.
- uncertainty: contains scripts to calculate and evaluate the uncertainty areas of the network and the manual labelling process.
GUIDE: guide.txt
TEST DATA: https://zenodo.org/record/4526867#.YCLd2-j0mbg
WEIGHTS: https://zenodo.org/record/4594326#.YEkro_4o8aw
TRAINED MODEL: https://zenodo.org/record/4526895#.YCLplOj0mbg
TRAINED MODEL FROZEN: https://zenodo.org/record/4594279#.YEkqJ_4o8aw
https://ieeexplore.ieee.org/document/8489861
If you benefit from this code or dataset, please cite our paper:
@article{Martin2018,
author={M. {Martin-Abadal} and E. {Guerrero-Font} and F. {Bonin-Font} and Y. {Gonzalez-Cid}},
journal={IEEE Access},
title={Deep Semantic Segmentation in an AUV for Online Posidonia Oceanica Meadows Identification},
year={2018},
volume={6},
number={},
pages={60956-60967},
keywords={autonomous underwater vehicles;image segmentation;neural nets;oceanographic techniques;seafloor phenomena;underwater equipment;diverse test sets;real-time semantic coverage maps;deep semantic segmentation;AUV;online Posidonia oceanica meadows identification;fundamental tools;deep neural network;highprecision semantic segmentation;sea-floor images;Semantics;Training;Image segmentation;Neural networks;Computer architecture;Decoding;Cameras;Deep learning;online semantic segmentation;Posidonia oceanica;autonomous underwater vehicle},
doi={10.1109/ACCESS.2018.2875412},
ISSN={2169-3536},
month={October},}
* If you use the network code, cite the source paper from the forked GitHub