MelBandRoFormerOptions
Options & Configuration · AiDotNet.Audio.SourceSeparation
Configuration options for the MelBand-RoFormer model.
Inheritance: ModelOptions → MelBandRoFormerOptions
For Beginners
This is an improved version of BS-RoFormer that divides frequencies using the mel scale (which matches how humans hear) instead of equal-width bands. This means more detail in the frequencies that matter most to us.
How It Works
MelBand-RoFormer (2024) extends BS-RoFormer by using mel-scale frequency bands instead of linear bands, better matching human perception. It achieves state-of-the-art SDR on vocals (13.2 dB) and other stems on the MUSDB18-HQ benchmark.
Constructors
| Constructor | Summary |
|---|---|
MelBandRoFormerOptions() |
Properties
| Property | Summary |
|---|---|
OnnxModelOptions OnnxOptions { get; set; } |
Gets or sets the ONNX Runtime settings used when the model is loaded from an ONNX file. |
bool UseRotaryEmbedding { get; set; } |
Gets or sets whether the paper’s rotary position embedding is used. |
double DropoutRate { get; set; } |
Gets or sets the dropout probability of the native layers during training. |
double LearningRate { get; set; } |
Gets or sets the learning rate of the default optimizer. |
double WeightDecay { get; set; } |
Gets or sets the weight decay of the default optimizer. |
int BandEmbeddingDim { get; set; } |
Gets or sets the width each frequency band is embedded to before the transformer. |
int FeedForwardDim { get; set; } |
Gets or sets the inner width of each feed-forward block. |
int FftSize { get; set; } |
Gets or sets the STFT window length in samples; the model rounds it up to a power of two. |
int HopLength { get; set; } |
Gets or sets the number of samples between successive STFT frames. |
int NumAttentionHeads { get; set; } |
Gets or sets the number of heads in each attention layer. |
int NumFreqBins { get; set; } |
Gets or sets the number of frequency bins per STFT frame (FftSize / 2 + 1). |
int NumMelBands { get; set; } |
Gets or sets the number of overlapping mel-scale bands the spectrogram is split into. |
int NumStems { get; set; } |
Gets or sets the number of stems the model separates. |
int NumTransformerLayers { get; set; } |
Gets or sets the number of transformer layers. |
int SampleRate { get; set; } |
Gets or sets the sample rate, in Hz, that input audio is expected at. |
int TransformerDim { get; set; } |
Gets or sets the hidden width of the transformer layers. |
string ModelPath { get; set; } |
Gets or sets the path of the pretrained ONNX model. |
string[] Sources { get; set; } |
Gets or sets the names of the separated stems, in output order. |