AudioEventDetectorOptions

Options & Configuration · AiDotNet.Audio.Classification

Options for audio event detection with Hershey et al. 2017’s audio ResNet-50: 96 x 64 log-mel patches of 960 ms (25 ms windows every 10 ms, 64 mel bins), a ResNet-50 whose first convolution has no stride, and a sigmoid output.

Inheritance: ModelOptions → AudioEventDetectorOptions

For Beginners

These options configure the AudioEventDetector model. Default values follow the paper (section 3).

Constructors

Constructor Summary
AudioEventDetectorOptions() Initializes a new instance with default values.
AudioEventDetectorOptions(AudioEventDetectorOptions other) Initializes a new instance by copying from another instance.

Properties

Property Summary
OnnxModelOptions OnnxOptions { get; set; } ONNX model options.
double Threshold { get; set; } Confidence threshold for event detection.
double WindowOverlap { get; set; } Window overlap ratio (0-1).
double WindowSize { get; set; } Window size in seconds.
int BaseChannels { get; set; } Channels of the first convolution and of the first bottleneck stage; each later stage doubles them and a bottleneck expands its stage width four times.
int FMax { get; set; } Maximum frequency for mel filterbank.
int FMin { get; set; } Minimum frequency for mel filterbank.
int FftSize { get; set; } STFT window in samples.
int HopLength { get; set; } Hop length.
int NumMels { get; set; } Number of mel bands.
int PatchFrames { get; set; } Spectrogram frames per classifier input.
int SampleRate { get; set; } Audio sample rate.
int[] StageDepths { get; set; } Bottleneck blocks per stage.
string ModelPath { get; set; } Path to ONNX model file (optional).
string[] CustomLabels { get; set; } Custom event labels (optional).