AffectiVR: Database for Periocular Identification and Valence Arousal Evaluation in Virtual Reality

AffectiVR is a comprehensive dataset designed for the identification of individuals through periocular recognition and the evaluation of emotional states in terms of valence and arousal within a virtual reality (VR) environment. This dataset aims to facilitate research and development in VR-based emotional analysis and biometric identification.

Authors

• Chaelin Seok
• Hyeji Lim
• Yeongje Park
• Junho Baek

Contact

• Email: chaelin9905@gmail.com

Dataset Download

The dataset can be downloaded via this link.

Dataset Structure

The structure of the AffectiVR dataset is illustrated below:

Example Data

Below is an example from the AffectiVR dataset:

Baseline Implementation

Preprocessing

The preprocessing step involves pupil detection, for which we use the model from Pupil Locator on GitHub.

Reflected Light Removal Example

Iris Code Generation Example

Model

The technical verification utilizes a model detailed in this IEEE paper. While the code is not publicly shared, requests for access can be considered via email.

Running the Baseline

Follow the steps outlined in this document to run the baseline models and evaluate the dataset.

Note

For additional information or to request code access, please contact us via the email provided above.