ppaudio 0.1.1

An audio classification toolkit based on PaddlePaddle for detecting abnormal sounds

ppaudio

A simple audio classification library based on PaddlePaddle.

Features

• Easy-to-use API for audio classification tasks
• Support for various audio classification scenarios:
  • Industrial noise detection
  • Fan noise classification
  • Engine sound analysis
  • Animal sound classification
  • Music genre classification
• Multiple feature extraction methods
• Various neural network architectures
• Visualization tools for audio analysis

Installation

pip install ppaudio

Quick Start

1. Prepare your data in CSV format:

   • train_data.csv
   • val_data.csv
   • test_data.csv

   Each CSV file should contain two columns: audio file path and label (0 or 1).

2. Create a configuration file (e.g., motor.yaml):

# Dataset parameters
dataset_conf:
  dataset:
    adjust_audio_length: True
    use_dB_normalization: True
    is_full_path: False
  sampler:
    batch_size: 64
    shuffle: True
    drop_last: True
  train_data: 'dataset/train_data.csv'
  val_data: 'dataset/val_data.csv'
  test_data: 'dataset/test_data.csv'

# Preprocessing parameters
preprocess_conf:
  feature_method: 'LogMelSpectrogram'
  method_args:
    sr: 48000
    n_fft: 1024
    hop_length: 512
    win_length: 1024
    window: 'hann'
    f_min: 50
    f_max: 14000
    n_mels: 64

# Loss function configuration
loss_conf:
  criterion: 'CrossEntropyLoss'

# Optimizer configuration
optimizer_conf:
  optimizer: 'Adam'
  optimizer_args:
    learning_rate: 0.001

# System configuration
sys_conf:
  use_GPU: True
  max_epoch: 60
  show_train_process: True
  save_train_process: False
  model_save_name: 'model'

3. Train your model:

import ppaudio

# Train model
ppaudio.train(config_path='motor.yaml')

4. Test your model:

# Test on dataset
ppaudio.test('model', 'dataset/test_data.csv')

# Test single audio file
ppaudio.test_single('model', '123.wav')

# Compare two audio files
ppaudio.compare('model', '123.wav', '456.wav', show_pic=True)

Documentation

For detailed documentation, please visit project documentation.

Examples

Here are some example use cases:

1. Industrial noise detection:

import ppaudio

# Train model
ppaudio.train(config_path='motor.yaml')

# Test new samples
ppaudio.test_single('model', 'new_sample.wav')

2. Music genre classification:

import ppaudio

# Train model (using a different configuration)
ppaudio.train(config_path='music.yaml')

# Compare two music samples
ppaudio.compare('model', 'rock.wav', 'jazz.wav', show_pic=True)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.