This repository contains Keras/Tensorflow code for the "CRF-RNN" semantic image segmentation method, implemented in TensorFlow Keras by https://github.com/sadeepj/crfasrnn_keras
Train: $ python train_gby.py -m fcn_RESNET50_8s -ds streets -e 1 -bs 32 -vb 1 -is 224
Eval/Test: $ python eval_gby.py -ds horsecoarse -w /storage/gby/semseg/horsecoarse_weights_fcn_RESNET50_8s_100ep
Parts of this work appear in Complex Relations in a Deep Structured Prediction Model for Fine Image Segmentation.
For citation:
@article{mata2018complex,
title={Complex Relations in a Deep Structured Prediction Model for Fine Image Segmentation},
author={Mata, Cristina and Ben-Yosef, Guy and Katz, Boris},
journal={arXiv preprint arXiv:1805.09462},
year={2018}
}