[Last update: September 16, 2024]
Period: 2023-10 - 2024-09
Status: completed
Author(s): Lucy Roellecke
Contact: lucy.roellecke[at]tuta.com
The GitHub Repository with all the code can be assessed here: github.com/lucyroe/AVR
The Code Documentation can be assessed here: lucyroe.github.io/AVR
Affective VR (AVR) aims to develop and test a tool for continous emotion ratings. The project proposes such a tool and assesses its effectiveness, usability and reliability using videos presented in virtual reality (VR).
The AVR project consists of three main stages: In the Selection Phase, three different rating methods ('Grid', 'Flubber' and 'Proprioceptive') were tested with different videos in VR of each 1 min length. In the Evaluation Phase, the 'Flubber' as winning rating method from the Selection Phase was tested for a longer VR experience of about 20 min with different videos playing after one another. In the Physio Phase, the very same stimuli and rating method are used but additionally to the behavioral data, EEG and periphysiological data are acquired.
This repository mainly focuses on the Physio Phase as it contains code I wrote in the framework of my master thesis with the title Embodied Emotion: Decoding dynamic affective states by integrating neural and cardiac data. However, this repository also contains code and results of earlier phase of the project by PhD Candidate Antonin Fourcade that I could use and profit off during the analyses for the thesis.
Code for all three phases can be found in ./code/AVR.
The whole preprocessing and analysis pipeline can be followed by running main.py.
The directory ./code/AVR/preprocessing contains scripts to preprocess annotation data and physiological data, respectively, of AVR Physio Phase.
The directory ./code/AVR/statistics contains scripts to calculate both univariate statistics, as well as modelling statistics.
The directory ./code/AVR/datacomparison contains scripts that compare continuous ratings from the Selection phase of the AVR project with the Physio Phase of the AVR project.
The directory ./code/AVR/modelling contains scripts to perform a Hidden Markov Model (HMM) analysis on data from the Physio phase to identify hidden affective states by using the physiological data.
The directory ./code/AVR/datavisualization contains plotting functions.
📂 code
├── 🐍 main.py
├── 📂 AVR
│ ├── 📁 datacomparison
│ ├── 📁 datavisualization
│ ├── 📁 statistics
│ ├── 📁 modelling
│ └── 📁 preprocessing
│ ├── 📁 annotation
│ └── 📁 physiological
└── 📁 configs
The results of all analyses can be found in ./results.
There is one sub-directory in the main results directory for phase of the AVR project: ./results/phase1, ./results/phase2, ./results/phase3, and ./results/comparison_phase1_phase2 and ./results/comparison_phase1_phase3 for comparing Selection and Evaluation/Physio phase.
📂 results
├── 📁 comparison_phase1_phase2
├── 📁 comparison_phase1_phase3
├── 📁 phase1
│ ├── 📁 assessment_results
│ ├── 📁 cocor_results
│ ├── 📁 cor_results
│ ├── 📁 cr_plots
│ ├── 📁 datacomparison
│ ├── 📁 datavisualization
│ ├── 📁 descriptives
│ └── 📁 icc_results
├── 📁 phase2
│ ├── 📁 cpa
│ └── 📁 descriptives
└── 📁 phase3
└── 📁 AVR
The figures published in my master thesis and the manuscripts can be found in ./publications.
📂 publications
├── 📁 articles
│ └── 📁 figures
└── 📁 thesis
└── 📁 figures
In case there is no project-related virtual / conda environment yet, create one for the project:
conda create -n AVR_3.11 python=3.11And activate it:
conda activate AVR_3.11Then install the code of the research project as python package:
# assuming your current working dircetory is the project root
pip install -e ".[develop]"Note: The -e flag installs the package in editable mode,
i.e., changes to the code will be directly reflected in the installed package.
Moreover, the code keeps its access to the research data in the underlying folder structure.
Thus, the -e flag is recommended to use.
Antonin Fourcade
Francesca Malandrone
Michael Gaebler
MIT