DannaSepOptions

Options & Configuration ยท AiDotNet.Audio.SourceSeparation

Configuration options for the Danna-Sep (Dual-path Attention Neural Network Audio Separator) model.

Inheritance: ModelOptions โ†’ DannaSepOptions

For Beginners

Danna-Sep separates mixed music into individual instruments by looking at audio in two ways: short-range patterns (what notes are being played right now) and long-range patterns (how the music evolves over time). This dual perspective helps it accurately pull apart overlapping instruments.

How It Works

Danna-Sep (2024) uses dual-path attention with interleaved intra-chunk and inter-chunk processing for music source separation. It achieves competitive results on MUSDB18 by efficiently modeling both local spectral patterns and long-range temporal dependencies.

Constructors

Constructor Summary
DannaSepOptions()

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } Gets or sets the ONNX runtime options.
double DropoutRate { get; set; } Gets or sets the dropout rate.
double LearningRate { get; set; } Gets or sets the learning rate.
int ChunkSize { get; set; } Gets or sets the chunk size for dual-path processing.
int EncoderDim { get; set; } Gets or sets the encoder dimension.
int FftSize { get; set; } Gets or sets the FFT size for STFT computation.
int HopLength { get; set; } Gets or sets the hop length for STFT computation.
int NumChannels { get; set; } Gets or sets the number of audio channels.
int NumDualPathBlocks { get; set; } Gets or sets the number of dual-path blocks.
int NumFreqBins { get; set; } Gets or sets the number of frequency bins (FftSize/2 + 1).
int NumHeads { get; set; } Gets or sets the number of attention heads.
int NumSources { get; set; } Gets or sets the number of sources to separate.
int SampleRate { get; set; } Gets or sets the audio sample rate in Hz.
string ModelPath { get; set; } Gets or sets the path to the ONNX model file.
string Variant { get; set; } Gets or sets the model variant.
string[] SourceNames { get; set; } Gets or sets the source names.