MadmomBeatTracker<T>

Models & Types · AiDotNet.Audio.MusicAnalysis

Madmom-style neural beat tracker using bidirectional RNNs.

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

Implements: IBeatTracker<T>

For Beginners

This model listens to music and finds exactly where each beat falls, like a musician tapping their foot in time. It can tell you the tempo (beats per minute) and mark every beat position, which is essential for music synchronization, DJ software, and automatic remixing.

Usage:

How It Works

The Madmom beat tracking system (Bock et al., 2016) uses a recurrent neural network to detect beat positions and downbeat positions in audio. It combines spectrogram features with bidirectional RNNs and a dynamic Bayesian network for beat tracking, achieving state-of-the-art results on multiple beat tracking benchmarks.

Constructors

Constructor Summary
MadmomBeatTracker(NeuralNetworkArchitecture<T> architecture, MadmomBeatTrackerOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) Creates a Madmom Beat Tracker in native training mode.
MadmomBeatTracker(NeuralNetworkArchitecture<T> architecture, string modelPath, MadmomBeatTrackerOptions options) Creates a Madmom Beat Tracker in native training mode.

Properties

Property Summary
double MaxBPM { get; }
double MinBPM { get; }

Methods

Method Summary
BeatTrackingResult<T> Track(Tensor<T> audio)
DownbeatResult<T> DetectDownbeats(Tensor<T> audio, BeatTrackingResult<T> beatTrackingResult)
IReadOnlyList<TempoHypothesis<T>> GetTempoHypotheses(Tensor<T> audio, int numHypotheses)
ModelMetadata<T> GetModelMetadata()
T EstimateTempo(Tensor<T> audio)
Task<BeatTrackingResult<T>> TrackAsync(Tensor<T> audio, CancellationToken cancellationToken)
Tensor<T> ComputeOnsetStrength(Tensor<T> audio)