Updates to DelphinID

.classpath

@@ -6,7 +6,7 @@
 			<attribute name="maven.pomderived" value="true"/>
 		</attributes>
 	</classpathentry>
-	<classpathentry kind="con" path="org.eclipse.jdt.launching.JRE_CONTAINER/org.eclipse.jdt.internal.debug.ui.launcher.StandardVMType/Amazon Coretto 21">
+	<classpathentry kind="con" path="org.eclipse.jdt.launching.JRE_CONTAINER/org.eclipse.jdt.internal.debug.ui.launcher.StandardVMType/jdk-21.0.2.13-hotspot">
 		<attributes>
 			<attribute name="module" value="true"/>
 			<attribute name="maven.pomderived" value="true"/>

pom.xml

@@ -310,7 +310,7 @@
 	<dependency>
 		<groupId>io.github.macster110</groupId>
 		<artifactId>jdl4pam</artifactId>
-		<version>0.0.98</version>
+		<version>0.0.99</version>
 	</dependency> 
 			
 	<!-- https://mvnrepository.com/artifact/gov.nist.math/jama -->

src/rawDeepLearningClassifier/dlClassification/delphinID/DelphinIDTest.java

@@ -228,7 +228,7 @@ public class DelphinIDTest {
 		String whistleContourPath = "D:/Dropbox/PAMGuard_dev/Deep_Learning/delphinID/testencounter415/whistle_contours.mat";
 
 		//the path to the model
-		String modelPath = "D:/Dropbox/PAMGuard_dev/Deep_Learning/delphinID/testencounter415/whistle_4s_encounter415.zip";
+		String modelPath = "D:/Dropbox/PAMGuard_dev/Deep_Learning/delphinID/testencounter415/whistle_model_2/whistle_4s_415.zip";
 		
 		//the path to the model
 		String matImageSave = "C:/Users/Jamie Macaulay/MATLAB Drive/MATLAB/PAMGUARD/deep_learning/delphinID/whistleimages.mat";
@@ -240,7 +240,7 @@ public class DelphinIDTest {
 		ArrayList<AbstractWhistleDataUnit> whistleContours = getWhistleContoursMAT(whistleContourPath);
 
 		//segment the whistle detections
-		ArrayList<SegmenterDetectionGroup> segments =  segmentWhsitleData(whistleContours,  dataStartMillis, 
+		ArrayList<SegmenterDetectionGroup> segments =  segmentWhsitleData(whistleContours,  (long) (dataStartMillis+(9.565*1000.)), 
 				segLen,  segHop);
 
 		for (int i=0; i<segments.size(); i++) {
@@ -267,7 +267,7 @@ public class DelphinIDTest {
 			float[] output =  predicition.get(0).getPrediction();
 		
 			System.out.println();
-			System.out.print("Segment: " +(aSegment.get(0).getSegmentStartMillis()-dataStartMillis)/1000.);
+			System.out.print("Segment: " + i + " " + (aSegment.get(0).getSegmentStartMillis()-dataStartMillis)/1000. + "s ");
 			for (int j=0; j<output.length; j++) {
 				System.out.print("  " + output[j]); 
 			}

src/rawDeepLearningClassifier/dlClassification/delphinID/DelphinIDWorker.java

@@ -12,6 +12,7 @@ import org.jamdev.jdl4pam.transforms.DLTransform.DLTransformType;
 import org.jamdev.jdl4pam.transforms.jsonfile.DLTransformsParser;
 import org.jamdev.jdl4pam.utils.DLMatFile;
 import org.jamdev.jdl4pam.utils.DLUtils;
+import org.jamdev.jpamutils.JamArr;
 import org.json.JSONArray;
 import org.json.JSONObject;
 
@@ -59,6 +60,7 @@ public class DelphinIDWorker extends  ArchiveModelWorker {
 		for (int i=0; i<dlParams.numClasses; i++) {
 			dlParams.binaryClassification[i]=true;
 		}
+	
 	}
 
 
@@ -94,6 +96,8 @@ public class DelphinIDWorker extends  ArchiveModelWorker {
 			}
 		}
 		
+	
+		
 		//something has gone wrong if we get here. 
 		return null; 
 	}
@@ -153,6 +157,8 @@ public class DelphinIDWorker extends  ArchiveModelWorker {
 		float[][][] transformedDataStack = new float[numChunks][][]; 
 
 		double[][] transformedData2; //spectrogram data
+		
+		
 		for (int j=0; j<numChunks; j++) {
 
 //			System.out.println("Number of whistle to process: " + whistleGroups.get(j).getStartSecond() + "s  " +  whistleGroups.get(j).getSubDetectionsCount() + "  " + whistleGroups.get(j).getSegmentStartMillis());
@@ -161,7 +167,7 @@ public class DelphinIDWorker extends  ArchiveModelWorker {
 			Whistles2Image whistles2Image = new Whistles2Image(whistleGroups.get(j), whistleImageParams);
 
 			//set the spec transform
-			((FreqTransform) this.getModelTransforms().get(0)).setSpecTransfrom(whistles2Image.getSpecTransfrom());
+			((FreqTransform) modelTransforms.get(0)).setSpecTransfrom(whistles2Image.getSpecTransfrom());
 
 			//process all the transforms. 
 			DLTransform transform = modelTransforms.get(0); 
@@ -170,6 +176,11 @@ public class DelphinIDWorker extends  ArchiveModelWorker {
 			}
 			
 			transformedData2 = ((FreqTransform) transform).getSpecTransfrom().getTransformedData(); 
+
+			//a bit ugly but works.
+			transformedData2 = JamArr.transposeMatrix(transformedData2);
+
+//			System.out.println("DelphinID input image: " + transformedData2.length + " x " + transformedData2[0].length  );
 			transformedDataStack[j] = DLUtils.toFloatArray(transformedData2); 
 			
 //			//TEMP

src/rawDeepLearningClassifier/dlClassification/delphinID/Whistles2Image.java

@@ -65,7 +65,7 @@ public class Whistles2Image extends FreqTransform {
 		ArrayList<double[][]> points = whistContours2Points(whistleGroup);
 
 		//does not work becaue it has to be on the AWT thread. 
-		BufferedImage canvas = makeScatterImage(points, size, new double[]{0, whistleGroup.getSegmentDuration()/1000.}, freqLimits,  5.);
+		BufferedImage canvas = makeScatterImage(points, size, new double[]{0, whistleGroup.getSegmentDuration()/1000.}, freqLimits,   10.);
 
 		double[][] imaged = new double[(int) size[0]][(int) size[1]];
 

src/test/rawDeepLearningClassifier/DelphinIDTest.java

@@ -4,6 +4,7 @@ import static org.junit.jupiter.api.Assertions.assertEquals;
 
 import java.awt.image.BufferedImage;
 import java.awt.image.Raster;
+import java.io.File;
 import java.io.IOException;
 import java.nio.file.Path;
 import java.nio.file.Paths;
@@ -14,8 +15,15 @@ import org.jamdev.jdl4pam.transforms.FreqTransform;
 import org.jamdev.jdl4pam.transforms.DLTransform.DLTransformType;
 import org.jamdev.jdl4pam.utils.DLMatFile;
 import org.jamdev.jdl4pam.utils.DLUtils;
+import org.jamdev.jpamutils.JamArr;
+import org.jamdev.jpamutils.interpolation.Bicubic;
+import org.jamdev.jpamutils.interpolation.Bilinear;
+import org.jamdev.jpamutils.interpolation.NearestNeighbor;
+import org.jamdev.jpamutils.spectrogram.SpecTransform;
 import org.junit.jupiter.api.Test;
 
+import ai.djl.MalformedModelException;
+import rawDeepLearningClassifier.dlClassification.archiveModel.SimpleArchiveModel;
 import rawDeepLearningClassifier.dlClassification.delphinID.Whistles2Image;
 import us.hebi.matlab.mat.format.Mat5;
 import us.hebi.matlab.mat.types.MatFile;
@@ -24,6 +32,55 @@ import us.hebi.matlab.mat.types.Matrix;
 public class DelphinIDTest { 
 	
 	
+	private static double[][] whistleScatter2Image(double[][] whistleValues) {
+		
+		//now perform the image transform in Java 
+		double[] freqLimits = new double[] {2000., 20000.};
+		double[] size = new double[] {680., 480.};
+		
+		ArrayList<double[][]> whistleImageArr = new ArrayList<double[][]>();
+		whistleImageArr.add(whistleValues);
+		
+		BufferedImage canvas = Whistles2Image.makeScatterImage(whistleImageArr, size, new double[]{50, 50. + 4.}, freqLimits,  6.);
+
+		double[][] imaged = new double[(int) size[0]][(int) size[1]];
+
+		float[] color = new float[3];
+		Raster raster = canvas.getData();
+		for (int i=0; i<imaged.length; i++) {
+			for (int j=0; j<imaged[0].length; j++) {
+				color = raster.getPixel(i,j, color);
+				imaged[i][j] = (255-color[0])/255.; //normalize
+			}
+		}
+		
+		//create the model transforms
+		ArrayList<DLTransform> modelTransforms = new ArrayList<DLTransform>();
+		modelTransforms.add(new FreqTransform(DLTransformType.SPECFLIP));
+//		modelTransforms.add(new FreqTransform(DLTransformType.SPECNORMALISE_MINIMAX));
+		modelTransforms.add(new FreqTransform(DLTransformType.SPECRESIZE, new Number[] {Integer.valueOf(60), Integer.valueOf(80), SpecTransform.RESIZE_BICUBIC}));
+		modelTransforms.add(new FreqTransform(DLTransformType.GAUSSIAN_FILTER, new Number[] {Double.valueOf(0.5)}));
+
+		
+		SpecTransform specTransform = new SpecTransform(); 
+		specTransform.setSpecData(imaged);
+		specTransform.setSampleRate((float) (freqLimits[1]*2)); 
+
+		
+		//set the spec transform
+		((FreqTransform) modelTransforms.get(0)).setSpecTransfrom(specTransform);
+
+		//process all the transforms. 
+		DLTransform transform = modelTransforms.get(0); 
+		for (int i =0; i<modelTransforms.size(); i++) {
+			transform = modelTransforms.get(i).transformData(transform); 
+		}
+		
+		double[][] transformedData2 = ((FreqTransform) transform).getSpecTransfrom().getTransformedData(); 
+		
+		return transformedData2;
+	}
+	
 	@Test
 	public void whistle2ImageTest() {
 	
@@ -44,22 +101,22 @@ public class DelphinIDTest {
 			//the values for the whistle detector.
 			double[][] whistleValues = DLMatFile.matrix2array(array);
 			
-			//the image after compression 
+//			//the image after compression 
-			array = matFile.getArray("image1compressedgrayscale");
+//			array = matFile.getArray("image1compressedgrayscale");
-			double[][] compressedWhistleImage = DLMatFile.matrix2array(array);
+//			double[][] compressedWhistleImage = DLMatFile.matrix2array(array);
-
+//
-			//the whistle2Image transform image
+//			//the whistle2Image transform image
-			array = matFile.getArray("image1originalgrayscalenorm");
+//			array = matFile.getArray("image1originalgrayscalenorm");
-			double[][] whislteImage = DLMatFile.matrix2array(array);
+//			double[][] whislteImage = DLMatFile.matrix2array(array);
 			
 			//now perform the image transform in Java 
-			double[] freqLimits = new double[] {0., 20000.};
+			double[] freqLimits = new double[] {2000., 20000.};
 			double[] size = new double[] {680., 480.};
 			
 			ArrayList<double[][]> whistleImageArr = new ArrayList<double[][]>();
 			whistleImageArr.add(whistleValues);
 			
-			BufferedImage canvas = Whistles2Image.makeScatterImage(whistleImageArr, size, new double[]{48, 48. + 4.}, freqLimits,  5.);
+			BufferedImage canvas = Whistles2Image.makeScatterImage(whistleImageArr, size, new double[]{50, 50. + 4.}, freqLimits,  10.);
 
 			double[][] imaged = new double[(int) size[0]][(int) size[1]];
 
@@ -67,33 +124,53 @@ public class DelphinIDTest {
 			Raster raster = canvas.getData();
 			for (int i=0; i<imaged.length; i++) {
 				for (int j=0; j<imaged[0].length; j++) {
-					color = raster.getPixel(i, j, color);
+					color = raster.getPixel(i,j, color);
 					imaged[i][j] = (255-color[0])/255.; //normalize
 				}
 			}
 			
-			ArrayList<DLTransform> transforms = new ArrayList<DLTransform>();
+			//create the model transforms
-			transforms.add(new FreqTransform(DLTransformType.SPECRESIZE, new Number[] {Integer.valueOf(64), Integer.valueOf(48)}));
+			ArrayList<DLTransform> modelTransforms = new ArrayList<DLTransform>();
+			modelTransforms.add(new FreqTransform(DLTransformType.SPECFLIP));
+//			modelTransforms.add(new FreqTransform(DLTransformType.SPECNORMALISE_MINIMAX));
+			modelTransforms.add(new FreqTransform(DLTransformType.SPECRESIZE, new Number[] {Integer.valueOf(60), Integer.valueOf(80), SpecTransform.RESIZE_BICUBIC}));
+			modelTransforms.add(new FreqTransform(DLTransformType.GAUSSIAN_FILTER, new Number[] {Double.valueOf(0.5)}));
 
-//			
-//			//set the spec transform
-//			((FreqTransform) transforms.get(0)).setSpecTransfrom(whistles2Image.getSpecTransfrom());
-//
-//			//process all the transforms. 
-//			DLTransform transform = modelTransforms.get(0); 
-//			for (int i =0; i<modelTransforms.size(); i++) {
-//				transform = modelTransforms.get(i).transformData(transform); 
-//			}
-//			
-//			transformedData2 = ((FreqTransform) transform).getSpecTransfrom().getTransformedData(); 
-//			transformedDataStack[j] = DLUtils.toFloatArray(transformedData2); 
-//			
 			
+			SpecTransform specTransform = new SpecTransform(); 
+			specTransform.setSpecData(imaged);
+			specTransform.setSampleRate((float) (freqLimits[1]*2)); 
+
+			
+			//set the spec transform
+			((FreqTransform) modelTransforms.get(0)).setSpecTransfrom(specTransform);
+
+			//process all the transforms. 
+			DLTransform transform = modelTransforms.get(0); 
+			for (int i =0; i<modelTransforms.size(); i++) {
+				transform = modelTransforms.get(i).transformData(transform); 
+			}
+			
+			double[][] transformedData2 = ((FreqTransform) transform).getSpecTransfrom().getTransformedData(); 
+			
+//			Bilinear interpolation1 = new Bilinear(JamArr.doubleToFloat(transformedData2));
+//			Bicubic interpolation2 = new Bicubic(JamArr.doubleToFloat(imaged));
+//
+//			System.out.println("Len input: " + imaged.length);
+//
+//			float[][] resizeArr = interpolation2.resize(Integer.valueOf(80), Integer.valueOf(60));
+//			
+//			System.out.println("Len resize: " + resizeArr.length);
+
+			System.out.println("Size Java: " + transformedData2.length + " x " + transformedData2[0].length);
 
 			//now save this image to a MATFILE
 			// Create MAT file with a scalar in a nested struct
 			MatFile matFileWrite = Mat5.newMatFile()
-			    .addArray("image1originalgrayscalenorm",DLMatFile.array2Matrix(imaged));
+			    .addArray("image1originalgrayscalenorm",DLMatFile.array2Matrix(imaged))
+			    .addArray("imagecompressedgrayscalenorm",DLMatFile.array2Matrix(transformedData2));
+//		    	.addArray("imagecompressedgrayscalenorm_nearest",DLMatFile.array2Matrix(JamArr.floatToDouble(resizeArr)));
+
 			// Serialize to disk using default configurations
 			Mat5.writeToFile(matFileWrite, "C:\\Users\\Jamie Macaulay\\MATLAB Drive\\MATLAB\\PAMGUARD\\deep_learning\\delphinID\\whistle_image_example_java.mat");
 			
@@ -105,9 +182,81 @@ public class DelphinIDTest {
 			assertEquals(false, false);
 		}
 
+	}
 	
+	@Test
+	public void modelInputTest() {
+		
+		System.out.println("DelphinID mode test start");
+
+		//ttest the model
+		String modelPath = "D:/Dropbox/PAMGuard_dev/Deep_Learning/delphinID/testencounter415/whistle_model_2/whistle_4s_415.zip";
+		
+		String relMatPath  =	"./src/test/resources/rawDeepLearningClassifier/DelphinID/whistle_image_example.mat";
+
+		try {
+			
+			//load the model
+			SimpleArchiveModel model = new SimpleArchiveModel(new File(modelPath));
+			
+
+			Path path = Paths.get(relMatPath);
+		
+			// Create MAT file with a scalar in a nested struct
+				MatFile matFile = Mat5.readFromFile(path.toString());
+				Matrix array = matFile.getArray("tfvalues");
+				
+				//the values for the whistle detector.
+				double[][] whistleValues = DLMatFile.matrix2array(array);
+				
+				//the image after compression 
+				array = matFile.getArray("image1compressedgrayscalenorm");
+				double[][] compressedWhistleImage = DLMatFile.matrix2array(array);
+				
+				//the raw whistle frequency time scatter values
+				array = matFile.getArray("tfvalues");
+				//the values for the whistle detector.
+				double[][] pamguardWhistleImage = 	whistleScatter2Image(whistleValues);
+	
+				
+				System.out.println("Size python: " + compressedWhistleImage.length + " x " + compressedWhistleImage[0].length);
+				
+				float[][][] input = new float[1][][];
+				input[0] =  JamArr.doubleToFloat(compressedWhistleImage);
+			
+				System.out.println("Model output: ");
+				float[] outputPython = model.runModel(input);
+				
+				input[0] =   JamArr.doubleToFloat(JamArr.transposeMatrix(pamguardWhistleImage));
+				
+				System.out.println("Size Java: " + 	input[0].length + " x " + 	input[0][0].length);
+
+				//a bit ugly but works.
+//				transformedData2 = JamArr.transposeMatrix(transformedData2);
+				
+				
+				float[] outputJava = model.runModel(input);
+				
+				for (int i=0; i<outputPython.length; i++) {
+					System.out.println(String.format("Output Python: %.4f Java: %.4f",outputPython[i],outputJava[i] )); 
+				}
+
+				MatFile matFileWrite = Mat5.newMatFile()
+					    .addArray("imagePython",DLMatFile.array2Matrix(compressedWhistleImage))
+					    .addArray("imageJava",DLMatFile.array2Matrix(JamArr.transposeMatrix(pamguardWhistleImage)));
+				Mat5.writeToFile(matFileWrite, "C:\\Users\\Jamie Macaulay\\MATLAB Drive\\MATLAB\\PAMGUARD\\deep_learning\\delphinID\\whistle_image_inout_test.mat");
+
+//			JamArr.printArray(output);
+			
+		} catch (MalformedModelException | IOException e) {
+			// TODO Auto-generated catch block
+			e.printStackTrace();
+		} 
+		
+		System.out.println("DelphinID mode test end");
 
 
 	}
 
+
 }