TRIOSlib

TRIOSlib is a library that contains implementations of state-of-the-art methods in Image Operator Learning. It has been used, in its many incarnations, in many research papers over the last 15 years. The current version (2.1) has had many parts rewritten and was redesigned to be flexible and to allow users to combine and extend each of its parts.

TRIOSlib can be installed using pip and the conda package/virtual environment manager.

pip install trios

Documentation: http://trioslib.github.io/

Building the documentation

The docs use mkdocs and the markdown-include extensions.

Both can be installed using pip

To view changes live in your browser run

$ mkdocs serve

and visit localhost:8000

To deploy run

$ mkdocs build -d trioslib.github.io

and commit/push changes in the submodule.

Notes for version 2.0.9

Before TRIOS 2.1 library versions were not specified in requirements.txt and setup.py. This lead to inconsistencies when loading operators saved with different versions of scikit-learn (specially sklearn.tree.DecisionTreeClassifier). Thus, we recommend using scikit-learn==0.17 when using operators trained with older versions of TRIOS.

We have created a tag v.2.0.10 for the modifications after release 2.0.9 but before 2.1. Some papers were written using TRIOSlib directly from the master branch and may require the use of this tag. It should be used with scikit-learn 0.17 as well.