Package deepWhistle

Class DeepWhistleMask

java.lang.Object

deepWhistle.DeepWhistleMask

All Implemented Interfaces:

PamFFTMask

public class DeepWhistleMask
extends Object
implements PamFFTMask

Field Summary

Fields

Modifier and Type	Field	Description
String	matFilePath

Constructor Summary

Constructors

Constructor	Description
DeepWhistleMask(DeepWhistleProcess maksedFFTProcess)

Method Summary

Modifier and Type	Method	Description
List<FFTDataUnit>	applyMask(List<FFTDataUnit> batch)	Apply the mask to a batch of FFTDataUnit objects.
boolean	initMask()	Initialize the mask.
static ai.djl.inference.Predictor<float[][],float[]>	loadPyTorchdeepWhistleModel(String modelPathS, long fftLen, long fftNum)	Load the deepWhistle into memory and create a predictor that can be called to run the model.
static float[][]	runPyTorchDeepWhislte(ai.djl.inference.Predictor<float[][],float[]> specPredictor, float[][] spectrogram)	Simple function which loads up the deep PyTorch whistle model and then runs it on completely random data

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

matFilePath

public String matFilePath

Constructor Details

DeepWhistleMask

public DeepWhistleMask(DeepWhistleProcess maksedFFTProcess)

Method Details

initMask

public boolean initMask()

Description copied from interface: PamFFTMask

Initialize the mask. Called once before processing starts.

Specified by:

initMask in interface PamFFTMask

Returns:

true if initialization was successful, false otherwise

applyMask

public List<FFTDataUnit> applyMask(List<FFTDataUnit> batch)

Description copied from interface: PamFFTMask

Apply the mask to a batch of FFTDataUnit objects. The mask performs some operation on the FFT data within each FFTDataUnit in the batch and returns the modified data units.

Specified by:

applyMask in interface PamFFTMask

Parameters:

batch - List of FFTDataUnit objects to process

Returns:

List of FFTDataUnit objects after applying the mask

loadPyTorchdeepWhistleModel

public static ai.djl.inference.Predictor<float[][],float[]> loadPyTorchdeepWhistleModel(String modelPathS,
 long fftLen,
 long fftNum)

Load the deepWhistle into memory and create a predictor that can be called to run the model.

Parameters:

modelPathS - - the path to the PyTorch model file

fftLen - - the fft length used in the model

fftNum - - the number of runs (i.e. number of FFT

Returns:

the predictor which returns a flattened confidence surface.

runPyTorchDeepWhislte

public static float[][] runPyTorchDeepWhislte(ai.djl.inference.Predictor<float[][],float[]> specPredictor,
 float[][] spectrogram)

Simple function which loads up the deep PyTorch whistle model and then runs it on completely random data