Speech Emotion Recognition Stress Tests

Testing suite for speech emotion recognition models.

It provides a command line tool to test the models for correctness, fairness, and robustness. Results are grouped after the different model tasks:

• arousal

• dominance

• valence

• emotional categories

🚨 Warning: you cannot install and run the test suite at the moment. It depends on an internal Python package to load all the models. Most of the datasets used in the tests are hosted on internal servers, and you would need to download the datasets from their original sources and convert to audformat to run the tests. We released the augmentation library used for the tests as the auglib Python package.