graph LR
denoiser_data_AudioDataset["denoiser.data.AudioDataset"]
denoiser_augment_Augment["denoiser.augment.Augment"]
denoiser_audio["denoiser.audio"]
denoiser_audio -- "provides services to" --> denoiser_data_AudioDataset
denoiser_data_AudioDataset -- "prepares data for" --> denoiser_augment_Augment
denoiser_augment_Augment -- "processes data from" --> denoiser_data_AudioDataset
The Data Management & Augmentation subsystem is responsible for preparing and augmenting audio datasets, feeding processed data to the model operations. It primarily involves the denoiser.data, denoiser.audio, and denoiser.augment modules.
This component acts as the central orchestrator of the data pipeline. Its core responsibility is to load raw audio files, match noisy and clean audio pairs, and structure them into a dataset suitable for consumption by the model's training or inference modules. It serves as the primary interface for data provisioning.
Related Classes/Methods:
This component is dedicated to applying various data augmentation techniques to the prepared audio data. Its role is critical for enhancing the diversity and robustness of the training dataset, which in turn helps improve the generalization capabilities of the model. It processes data received from the AudioDataset.
Related Classes/Methods:
This module functions as a foundational utility layer, providing low-level operations for interacting with audio files. It is responsible for discovering audio files within the file system and extracting essential metadata required for data loading and pairing. It supports the AudioDataset by providing the raw information needed.
Related Classes/Methods: