Based on Keras/Tensorflow implementation here: https://github.com/sadeepj/crfasrnn_keras, implemented for the following paper:
@inproceedings{crfasrnn_ICCV2015,
author = {Shuai Zheng and Sadeep Jayasumana and Bernardino Romera-Paredes and Vibhav Vineet and
Zhizhong Su and Dalong Du and Chang Huang and Philip H. S. Torr},
title = {Conditional Random Fields as Recurrent Neural Networks},
booktitle = {International Conference on Computer Vision (ICCV)},
year = {2015}
}
Their live demo: http://crfasrnn.torr.vision
Their Caffe version: http://github.com/torrvision/crfasrnn
Paper link: http://www.robots.ox.ac.uk/~szheng/papers/CRFasRNN.pdf
python train.py -m <model_name> -ds <dataset_name> -is <input_size> -e <num_epochs> -bs <batch_size> -vb 1 -g <gpu_number>
example
python train.py -m fcn_RESNET50_8s -ds voc2012 -is 224 -e 1 -bs 32 -vb 1 -g 0