README.md

metadata

configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
dataset_info:
  features:
    - name: image
      dtype: image
    - name: label
      dtype:
        class_label:
          names:
            '0': afro
            '1': classical
            '2': country
            '3': disco
            '4': electro
            '5': jazz
            '6': latin
            '7': metal
            '8': pop
            '9': rap
            '10': reggae
            '11': rock
  splits:
    - name: train
      num_bytes: 128963338.5857826
      num_examples: 1697
    - name: test
      num_bytes: 14256351.565217393
      num_examples: 189
  download_size: 143291941
  dataset_size: 143219690.151
license: mit
task_categories:
  - image-classification
tags:
  - music
size_categories:
  - 1K<n<10K

Dataset Card

The egtzan_plus dataset is an GTZAN like dataset for musical genre classification in the vision domain. In egtzan_plus, new classes such as Electro and Afro have been added to the original GTZAN dataset. Each audio track (30s) is transformed into a Mel-frequency spectrogram using Librosa:

# Mel-frequency spectrogram generation
y, sr = librosa.load(audio_file)
ms = librosa.feature.melspectrogram(y=y, sr=sr, n_mels=128, fmax=8000)
log_ms = librosa.power_to_db(ms, ref=np.max)
librosa.display.specshow(log_ms)

The dataset contains the following classes:

• Afro
• Classical
• Country
• Disco
• Electro
• Jazz
• Latin
• Metal
• Pop
• Rap
• Reggae
• Rock

The dataset is split into train and test sets as follows:

• Train: 1697 examples
• Test: 189 examples