Fully convolutional networks for building and road extraction

This is the reference implementation of the codes and models for the FCN-4s in the [IGARSS2016 FCN-4s paper](paper/FCN for building and road extraction.pdf).

Fully convolutional networks for building and road extraction: Preliminary results
Zilong Zhong*, Jonathan Li, Weihong Cui, Han Jiang
IGARSS 2016

Futher exploration about FCN models please refer to shelhamer/fcn.berkeleyvision.org.

Fully Convolutional Models for Semantic Segmentation
Jonathan Long*, Evan Shelhamer*, Trevor Darrell
CVPR 2015
arXiv:1411.4038

Before deploying this model, make sure you have installed the latest BVLC/caffe:master.
FCN-4s Model for building extraction in high spatial resolution imagery.

Training command:

python TrainingVal/solve.py

Evaluating command:

python evaluate.py