Personalization using enrollment

This repository contains the code needed to reproduce the experiments in (Triantafyllopoulos and Schuller, 2024).

List of python files

1. main.py: code used to start training, utilizing hydra configuration files in configs
2. models.py: code to create models
3. data.py: implementation of datasets
4. evaluate.py: code to compute Gini index and CIs
5. iswf.py: code to create ISWF plots
6. predict_test_msp.py: code to evalute on MSP test set
7. speaker_plot.py: code to plot speaker-level UAR (not used in paper)
8. training.py: code implementing training

Usage

1. Download MSP-Podcast and FAU-AIBO manually. To do this, you need an EULA with the dataset owners.
2. Run main.py. You will be asked to provide the root for the data and the results-root to store your results.
3. (Optional) Run evaluate.py and iswf.py to run the detailed evaluations of the paper.

Adaptation data

Adaptation CSVs are included under adaptation-sets for each dataset/task. Only filenames are included (need to request datasets from respective owners).

Reference

Triantafyllopoulos, A., Schuller, B., (2024), "Enrolment-based personalisation for improving individual-level fairness in speech emotion recognition," Proc. INTERSPEECH, Kos Island, Greece, (accepted).

@inproceedings{Triantafyllopoulos24-EPF
    author={Triantafyllopoulos, Andreas, and Schuller, Björn},
    title={Enrolment-based personalisation for improving individual-level fairness in speech emotion recognition},
    year={2024},
    booktitle={Proc. INTERSPEECH},
    address={Kos Island, Greece}
}