EyeFeatures is an open-source Python package for analyzing eye movement data in any visual task. Its capabilities encompass preprocessing, visualization, statistical analysis, feature engineering and machine learning. Its unique feature is its architecture and versatility. Accepting data in .csv format containing gaze position coordinates, the package allows filtration of raw data to remove noise and detecting fixations and saccades with different algorithms. Having fixations any standard descriptive statistical eye movement features (such as totalFD, meanFD etc.) can be computed, including AOI-wise features. AOIs can be predefined or assigned automatically. More complex features, such as chaos measures, topological features, density maps, scanpath similarities for various distance metrics can be computed as well. The package allows to account for the panel structure of the data, calculating shift features relative to group averages. The visualization module allows output a variety of visualization options, including static and dynamic scanpath plots. The architecture of the package allows seamless embedding of its preprocessing and feature extraction classes in Sklearn pipelines. Moreover, it provides datasets and models for deep learning with Pytorch.
It is recommended to install package in separate python environment: (If you want to install it to base environment, ingore steps 1-2)
- In conda you can create it with
conda create -n <name_of_environment> - To activate environment write
conda activate <name_of_environment>. In order to make it visible in jupyter writepip install ipykernelandpython -m ipykernel install --user --name <name_of_environment> --display-name "<name_of_environment>
By default eyefeatures is installed without deep module:
- To install eyefeatures write
pip install eyefeatures.
If you want to install it with deep module:
- Write command
git clone https://github.com/hse-scila/EyeFeatures(in windows you need to do it in anaconda prompt). - Write command
cd EyeFeatures. - Write command
pip install poetry. - Write
poetry install --with deep.
Documentation for the latest version can be found here. Documentation contains description of all classes, functions and their parameters.
You can find notebooks with tutorials devoted to differnet parts of the library in this reposiry in tutorial folder.
Extensive table with references to all methods is coming soon