Installation

The following installation steps were tested with Debian 9 (stretch), Ubuntu 18.04 (Bionic Beaver) and Ubuntu 16.04 (Xenial Xerus). Please, run all commands either as root or sudoer user.

1) To install python3 and some dependencies, one must run, in terminal:

apt update && apt install -y python3 python3-gdal python3-pip python3-dev wget

2) Install TensorFlow, Keras and scikit-learn:

pip3 install tensorflow keras sklearn

3) Install RIOS:

wget https://bitbucket.org/chchrsc/rios/downloads/rios-1.4.5.tar.gz && tar -xvzf rios-1.4.5.tar.gz && cd rios-1.4.5 && python3 setup.py install --prefix=/opt/rios-1.4.5 && export PATH=$PATH:/opt/rios-1.4.5/bin/ && export PYTHONPATH=/opt/rios-1.4.5/lib/python$(python3 --version | cut -c8-10)/site-packages/ && cd .. && rm -rf rios-1.4.5 rios-1.4.5.tar.gz

Usage Example

First, some dataset has to be grabbed for the training to work. You can download an AVIRIS example scene issuing the following command:

wget https://www.lapig.iesa.ufg.br/drive/index.php/s/IWebcJhmreFeZYP/download -O data/example_scene.tar.gz tar -xvzf example_scene.tar.gz Default data directory is ./data, but you can place a custom directory in the optional [data_dir].

Training the model

python3 run.py train [data_dir]

Evaluating the model

python3 run.py eval [data_dir]

Running predict over the entire scene

python3 run.py predict [data_dir]

EvandroCT/dl-generalize