VR-Dataset-Emotions

This repository is for our paper published in TVCG "An Immersive and Interactive VR Dataset to Elicit Emotions".

This dataset includes:

• Five VR scenes that can elicit different emotions.
• An example VR app that includes all the five scenes, including an interactive SAM qestionnaire scene.
• [Optional] A server that can record the user's behavior and SAM reports.

Scene Modeling

Our VR scenes are modeled based on the original validated 360° video scenes. The modeling process includes:

1. Scene Selection: Select the scenes that can elicit the targeted emotions.
2. Scene Modeling: Model the scenes in Unity and Blender with the same content and layout as the original 360° video scenes.
3. Lighting and Texturing: Adjust the lighting and texturing to make the scenes more immersive.
4. Interaction Design: Enable users to be able to teleport to explore the scenes.

An example comparison between the original 360° video scene (left) and the modeled VR scene (right) is shown below:

Targeted Emotions

Our dataset has been validated by 160 participants (80 participants for modeled VR scenes, 80 participants for the original 360° video scenes). The mean values of the targeted emotions are shown below:

For more details about the dataset, please refer to our paper here: (to be published)

Environment Requirements

• Unity 2021.3 Long-Term-Support (LTS) version.
• OpenXR plugin for Unity.
• VR headset: the VR app was tested on Oculus Quest 2, you may need to modify the settings for other VR devices.
• [Optional] Python 3 with Django.

Quick Start

Test VR Scenes

1. Download or clone this repository.
2. Load the Unity project in the unity folder.
3. For each scene, open the scene file in the Scenes folder and press the play button to test the scene.
4. [Optional] Build and deploy the VR app to your VR headset:
   • Go to File -> Build Settings -> Build to build the app.
   • Follow the instructions to deploy the app to your VR headset.

[Optional] Remote Server

The remote server is for recording user's behavior and SAM reports in VR using Django. You need to install Django to run the server:

pip install numpy django django-cors-headers

It is recommended to use a virtual environment such as venv or Anaconda.

For quick test, go to the server folder and run the server.

You may need to create a superuser to access the admin page:

python manage.py changepassword admin

Then run the server:

python manage.py runserver

Then go to 127.0.0.1:8000/admin in your browser to check the recorded data.

For detailed data recording structure, please refer to the server/emotion_app/models.py file.

[Optional] Unity Settings for Data Recording

Recording SAM Reports

1. Open the Assets/Scenes/EmotionSurvey scene.
2. Go to the SurveySAM object in the hierarchy.
3. In the SAM Survey Events script component, set the Server URL to your server URL.

• For local test, you can set it to http://127.0.0.1:8000/emotion-survey/.

Recording User's Behavior

1. Open the corresponding scene in the Assets/Scenes folder.
2. Go to the SceneController object in the hierarchy.
3. In the Camera Post Sender script component, set the Server URL to your server URL.

• For local test, you can set it to http://127.0.0.1:8000/camera-post/.
• Alternatively, open the Scripts/Utility/CameraPostSender.cs script and modify the serverURL variable, this will become the default setting in the Unity editor for all scenes.

For actual deployment, follow the instructions in the Django documentation.

Citation

If you find this dataset useful, please cite our paper:

@ARTICLE{jiang2024immersive,
    author = {Weiwei Jiang and Maximiliane Windl and Benjamin Tag and Zhanna Sarsenbayeva and Sven Mayer},
    journal = {IEEE Transactions on Visualization & Computer Graphics},
    title = {An Immersive and Interactive VR Dataset to Elicit Emotions},
    year = {2024},
    volume = {},
    number = {01},
    issn = {1941-0506},
    pages = {1-11},
    doi = {10.1109/TVCG.2024.3456202},
    publisher = {IEEE Computer Society},
    address = {Los Alamitos, CA, USA},
    month = {sep}
}