alantanlc / torchemotion

Emotion recognition library for PyTorch

torchemotion: an emotion recognition library for PyTorch

The aim of torchemotion is to apply PyTorch and torchaudio to the emotion recognition domain. We begin with providing basic dataloaders to read popular emotional datasets.

• Dataloaders for common emotional speech datasets

Dataloaders

Dataloaders for the following emotional datasets are currently available:

• IEMOCAP - Interactive Emotional Dyadic Motion Capture
• EmoDB - Berlin Database of Emotional Speech
• RAVDESS - Ryerson Audio-Visual Database of Emotional Speech and Song

Example Usage

IEMOCAP

First, obtain the IEMOCAP dataset by following the steps here. You will need to submit an electronic release form to SAIL lab at USC. They respond pretty quickly (3-5 days) and will email you the download link.

Unzip the file after download.

Then, initialize an IemocapDataset object by passing in the path to the IEMOCAP_full_release directory.

import IemocapDataset

# Initialize IemocapDataset
iemocap_dataset = IemocapDataset('./IEMOCAP_full_release')

# Iterate over data
for index, sample in enumerate(iemocap_dataset):
    print(index, sample)

EmoDB

First, download and unzip the EmoDB dataset from here.

Then, initialize an EmodbDataset object by passing in the path to the download directory.

import EmodbDataset

# Initialize EmodbDataset
emodb_data = EmodbDataset('./download')

# Iterate over data
for index, sample in enumerate(emodb_dataset):
    print(index, sample)

RAVDESS

First, download and unzip the RAVDESS dataset from here.

Then, initialize an RavdessDataset object by passing in the path to the Audio_Speech_Actors_01-24 directory.

import RavdessDataset

# Initialize RavdessDataset
ravdess_dataset = RavdessDataset('./Audio_Speech_Actors_01-24')

# Iterate over data
for index, sample in enumerate(ravdess_dataset):
    print(index, sample)

Dependencies

• pytorch
• torchaudio
• pandas
• numpy

Disclaimer on Datasets

This library does not host or distribute these dataset, or claim that you have license to use the dataset. It is your responsiblity to determine whether you have permission to use the dataset under the dataset's license.

If you're a dataset owner and wish to update any part of it (description, citation, etc.), or do not want your dataset to be included in this library, please get in touch through a GitHub issue. Thanks for your contribution to the speech emotion recognition community!