ghermoso/egtzan_plus

---
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:

```python
# 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

原始信息汇总

数据集概述

数据集信息

• 名称: egtzan_plus
• 描述: 用于音乐流派分类的GTZAN类数据集，新增了Electro和Afro等类别。每个音频轨道（30秒）通过Librosa转换为Mel频率谱图。

数据结构

• 特征:
  • image: 图像类型，表示Mel频率谱图。
  • label: 类别标签，包含以下类别:
    • Afro
    • Classical
    • Country
    • Disco
    • Electro
    • Jazz
    • Latin
    • Metal
    • Pop
    • Rap
    • Reggae
    • Rock

数据分割

• 训练集:
  • 样本数: 1697
  • 大小: 128963338.5857826字节
• 测试集:
  • 样本数: 189
  • 大小: 14256351.565217393字节

数据集大小

• 下载大小: 143291941字节
• 数据集大小: 143219690.151字节

许可

• 许可证: MIT

任务类别

• 图像分类

标签

• 音乐

大小类别

• 1K<n<10K