ConformerFP<T>

Models & Types · AiDotNet.Audio.Fingerprinting

Conformer-based audio fingerprinting model combining self-attention with convolutions.

Inheritance: NeuralNetworkBase<T> → AudioNeuralNetworkBase<T> → ConformerFP<T>

Implements: IAudioFingerprinter<T>

For Beginners

ConformerFP uses a powerful architecture that combines two ways of understanding audio: attention (looking at the big picture) and convolutions (looking at local details). This combination makes it very good at creating fingerprints that can identify songs even from noisy or distorted recordings.

Usage:

How It Works

ConformerFP applies the Conformer architecture (convolution-augmented Transformer) to audio fingerprinting. It combines self-attention for global context with convolutions for local feature extraction, producing highly robust fingerprints for large-scale audio retrieval.

Constructors

Constructor Summary
ConformerFP(NeuralNetworkArchitecture<T> architecture, ConformerFPOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) Creates a ConformerFP model in native training mode.
ConformerFP(NeuralNetworkArchitecture<T> architecture, string modelPath, ConformerFPOptions options) Creates a ConformerFP model in native training mode.

Properties

Property Summary
int FingerprintLength { get; }
string Name { get; }

Methods

Method Summary
AudioFingerprint<T> Fingerprint(Tensor<T> audio)
AudioFingerprint<T> Fingerprint(Vector<T> audio)
IReadOnlyList<FingerprintMatch> FindMatches(AudioFingerprint<T> query, AudioFingerprint<T> reference, int minMatchLength)
ModelMetadata<T> GetModelMetadata()
double ComputeSimilarity(AudioFingerprint<T> fp1, AudioFingerprint<T> fp2)