GraFPrintOptions

Options & Configuration · AiDotNet.Audio.Fingerprinting

Configuration options for the GraFPrint graph-based audio fingerprinting model.

Inheritance: ModelOptions → GraFPrintOptions

For Beginners

GraFPrint treats a song’s spectrogram as a network (graph) of connected sound points, then uses a special AI called a graph neural network to turn that network into a fingerprint. This approach captures how different parts of the sound relate to each other, making it very robust to distortions like noise or tempo changes.

How It Works

GraFPrint uses graph neural networks to model spectro-temporal relationships in audio for robust fingerprinting. It constructs a graph from spectrogram features where nodes represent time-frequency points and edges capture local relationships, then applies GNN layers to produce compact fingerprint embeddings.

Constructors

Constructor Summary
GraFPrintOptions()

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } Gets or sets the ONNX runtime options.
bool DisableFusedOptimizerStep { get; set; } Legacy compatibility switch.
double DropoutRate { get; set; } Gets or sets the maximum stochastic-depth rate across graph blocks.
double LearningRate { get; set; } Adam learning rate from the official GraFPrint configuration.
double MatchThreshold { get; set; } Gets or sets the cosine similarity threshold for considering a fingerprint match.
double MaxGradNorm { get; set; } Optional global L2 gradient clipping threshold.
double MinimumLearningRate { get; set; } Gets or sets the cosine scheduler’s minimum learning rate.
double SegmentDurationSec { get; set; } Gets or sets the segment duration in seconds.
double Temperature { get; set; } Gets or sets the contrastive loss temperature.
int EmbeddingDim { get; set; } Gets or sets the fingerprint embedding dimension.
int EncoderEmbeddingDim { get; set; } Gets or sets the graph encoder readout width before the SimCLR projector.
int FftSize { get; set; } Gets or sets the FFT window size.
int GnnHiddenDim { get; set; } Gets or sets the GNN hidden dimension.
int HopLength { get; set; } Gets or sets the hop length between frames.
int KNeighbors { get; set; } Gets or sets the k-nearest neighbors for graph construction.
int LRSchedulerTMax { get; set; } Maximum step count for cosine annealing.
int NumAttentionHeads { get; set; } Gets or sets the number of graph attention heads.
int NumGnnLayers { get; set; } Gets or sets the number of GNN layers.
int NumMels { get; set; } Gets or sets the number of mel filterbank channels.
int PeakFilters { get; set; } Gets or sets the learned peak-extractor output channels.
int PeakKernelSize { get; set; } Gets or sets the square peak-extractor convolution kernel size.
int PeakStride { get; set; } Gets or sets the frequency-only stride in the learned peak extractor.
int ProjectionExpansion { get; set; } Gets or sets the projector hidden-width multiplier relative to the fingerprint size.
int SampleRate { get; set; } Gets or sets the expected audio sample rate in Hz.
string ModelPath { get; set; } Gets or sets the path to the ONNX model file.