SER Odyssey Baseline WavLM Dominance

🚀 WavLM-based Audio Classification Model for Emotion Recognition

This model is designed for audio classification, specifically for emotion recognition. It was trained on the MSP-Podcast dataset for the Odyssey 2024 Emotion Recognition competition baseline. The single - task specialized dominance model predicts dominance within a range of approximately 0...1.

📚 Documentation

Model Information

Benchmarks

The following table shows the CCC (Concordance Correlation Coefficient) based on the Test3 and Development sets of the Odyssey Competition in a single - task setup:

For more details, you can refer to the demo, paper and GitHub.

Citation

@InProceedings{Goncalves_2024,
            author={L. Goncalves and A. N. Salman and A. {Reddy Naini} and L. Moro-Velazquez and T. Thebaud and L. {Paola Garcia} and N. Dehak and B. Sisman and C. Busso},
            title={Odyssey2024 - Speech Emotion Recognition Challenge: Dataset, Baseline Framework, and Results},
            booktitle={Odyssey 2024: The Speaker and Language Recognition Workshop)},
            volume={To appear},
            year={2024},
            month={June},
            address =  {Quebec, Canada},
}

💻 Usage Examples

Basic Usage

from transformers import AutoModelForAudioClassification
import librosa, torch

#load model
model = AutoModelForAudioClassification.from_pretrained("3loi/SER-Odyssey-Baseline-WavLM-Dominance", trust_remote_code=True)

#get mean/std
mean = model.config.mean
std = model.config.std


#load an audio file
audio_path = "/path/to/audio.wav"
raw_wav, _ = librosa.load(audio_path, sr=model.config.sampling_rate)

#normalize the audio by mean/std
norm_wav = (raw_wav - mean) / (std+0.000001)

#generate the mask
mask = torch.ones(1, len(norm_wav))

#batch it (add dim)
wavs = torch.tensor(norm_wav).unsqueeze(0)


#predict
with torch.no_grad():
    pred = model(wavs, mask)

print(model.config.id2label) 
print(pred)
#{0: 'dominance'}
#tensor([[0.3670]])

Advanced Usage

The provided basic usage example covers the main process of using the model for prediction. For more advanced scenarios, you may need to adjust the pre - processing steps according to your specific requirements, such as using different normalization methods or handling different audio formats. The core prediction process remains the same as shown above.