This is a collabrative repositary hosting code to identify channel pixels from a flume experiment image, with optionally additional data like elevation maps.
Example data sets are provided in data/test_data
. The images in the folder are in groups of three: one correspond to the mask (target for the model), one correspond to the RGB image, and another one correspond to the elevation model (optional data). For example:
The module data_preparing
provides methods to generate the groups of images from a data source like numpy.ndarray
or matlab's .mat
files.
We recommend using virtual environments to isolate the program environment. The file requirements.txt
contains a list of module needed for the project. To setup an environment using python3-venv
:
cd /path/to/repo
python3 -m venv .venv
source .venv/bin/active
pip3 install -r requirements.txt
This project uses Tensorflow to construct the CNN models.
GPU or better resource is recommended to train the model. One could experiment or demo with a small sample dataset on a CPU only environment.