/FCN-CD-PyTorch

Training code for https://github.com/rcdaudt/fully_convolutional_change_detection

Primary LanguagePythonBSD 2-Clause "Simplified" LicenseBSD-2-Clause

Fully Convolutional Siamese Networks for Change Detection

This is an unofficial implementation of the paper

Rodrigo Caye Daudt, Bertrand Le Saux, Alexandre Boulch. (2018, October). Fully convolutional siamese networks for change detection. In 2018 25th IEEE International Conference on Image Processing (ICIP) (pp. 4063-4067). IEEE.

as the official repo does not provide the training code.

paper link

Basic Usage

# The network definition scripts are from the original repo
git clone --recurse-submodules git@github.com:Bobholamovic/FCN-CD-PyTorch.git

For training, try

# In the root directory of this repository
mkdir exp
cd src
python train.py train --exp-config ../config_base.yaml

For evaluation, try

python train.py val --exp-config ../config_base.yaml --resume path_to_checkpoint

You can find the checkpoints in exp/base/weights/, the log files in exp/base/logs, and the output change maps in exp/outs.