Skip to main content

Dataset Preparation

Overview

The Dataset Manager module is used to import, organize, preview, and prepare audio datasets for model training.
It supports WAV, MP3, and FLAC formats and provides tools to manage dataset structure efficiently.

Purpose

The Dataset Manager allows users to:

  • Import large collections of audio files
  • Organize data by class or label
  • Preview audio using waveform and spectrogram views
  • Move selected data into the training dataset structure

Supported Audio Formats

  • WAV – Preferred format for training
  • MP3 – Supported for dataset import
  • FLAC – Supported for dataset import

Note:

Although MP3 and FLAC formats are supported during import, it is strongly recommended to convert all audio files to WAV format before training. WAV files preserve uncompressed audio data and provide consistent feature extraction.