Adapt to changes in KNIME Deep Learning

Author: dietzc

org.knime.knip.dl/src/org/knime/ip/dl/DLKnipUtil.java

@@ -1,6 +1,5 @@
 package org.knime.ip.dl;
 
-import org.knime.dl.core.DLDefaultDimensionOrder;
 import org.knime.dl.core.DLDimension;
 import org.knime.dl.core.DLDimensionOrder;
 import org.knime.dl.core.DLTensorSpec;
@@ -17,71 +16,80 @@ final class DLKnipUtil {
 	private DLKnipUtil() {
 		// utility class
 	}
-	
-	public static <T extends RealType<T>> RandomAccessibleInterval<T> mapImgToDL(ImgPlusValue<T> img, DLTensorSpec tensorSpec) {
+
+	public static <T extends RealType<T>> RandomAccessibleInterval<T> mapImgToDL(ImgPlusValue<T> img,
+			DLTensorSpec tensorSpec) {
 		int[] mapping = calculateMapping(img, tensorSpec);
 		return DimSwapper.swap(img.getImgPlus(), mapping);
 	}
-	
+
 	private static DLDimensionOrder extractDimensionOrder(DLTensorSpec tensorSpec) {
-		if (tensorSpec.getDimensionOrder() == DLDefaultDimensionOrder.Unknown) {
+		if (tensorSpec.getDimensionOrder() == DLDimensionOrder.Unknown) {
 			throw new IllegalArgumentException(
 					"Can't infer shape from image if the dimension order of the input tensor is unknown");
 		}
 		return tensorSpec.getDimensionOrder();
 	}
-	
-	public static <T extends RealType<T>> long[] getShapeFromImg(final ImgPlusValue<T> img, final DLTensorSpec tensorSpec) {
+
+	public static <T extends RealType<T>> long[] getShapeFromImg(final ImgPlusValue<T> img,
+			final DLTensorSpec tensorSpec) {
 		int[] mapping = calculateMapping(img, tensorSpec);
 		return mapShape(img.getDimensions(), mapping);
 	}
-	
+
 	private static <T extends RealType<T>> int[] calculateMapping(final ImgPlusValue<T> img, DLTensorSpec tensorSpec) {
 		DLDimensionOrder tensorDimensionOrder = extractDimensionOrder(tensorSpec);
 		DLDimension[] imgDimensionOrder = getDimensionOrder(getAxes(img.getMetadata()));
 		return tensorDimensionOrder.inferMappingFor(imgDimensionOrder);
 	}
-	
+
 	private static long[] mapShape(final long[] imgShape, final int[] mapping) {
 		assert imgShape.length == mapping.length;
 		long[] mappedShape = new long[imgShape.length];
 		for (int i = 0; i < mappedShape.length; i++) {
 			// in KNIP the last dimension changes the slowest (e.g. C in XYC) while
-			// in deep learning (especially TensorFlow) the last dimension changes the fastest.
+			// in deep learning (especially TensorFlow) the last dimension changes the
+			// fastest.
 			mappedShape[i] = imgShape[imgShape.length - mapping[i] - 1];
 		}
 		return mappedShape;
 	}
-	
+
 	private static DLDimension[] getDimensionOrder(CalibratedAxis[] axes) {
 		DLDimension[] dimOrder = new DLDimension[axes.length];
 		for (int i = 0; i < axes.length; i++) {
 			// in KNIP the last dimension changes the slowest (e.g. C in XYC) while
-			// in deep learning (especially TensorFlow) the last dimension changes the fastest.
+			// in deep learning (especially TensorFlow) the last dimension changes the
+			// fastest.
 			dimOrder[i] = axisToDimension(axes[axes.length - i - 1]);
 		}
 		return dimOrder;
 	}
-	
+
 	private static CalibratedAxis[] getAxes(final ImgPlusMetadata metaData) {
 		CalibratedAxis[] axes = new CalibratedAxis[metaData.numDimensions()];
 		metaData.axes(axes);
 		return axes;
 	}
-	
+
 	private static DLDimension axisToDimension(CalibratedAxis axis) {
 		switch (axis.type().getLabel()) {
-		case "X": return DLDimension.Width;
-		case "Y": return DLDimension.Height;
-		case "Z": return DLDimension.Depth;
-		case "Channel": return DLDimension.Channel;
-		case "Time": return DLDimension.Time;
+		case "X":
+			return DLDimension.Width;
+		case "Y":
+			return DLDimension.Height;
+		case "Z":
+			return DLDimension.Depth;
+		case "Channel":
+			return DLDimension.Channel;
+		case "Time":
+			return DLDimension.Time;
 
 		default:
 			throw new IllegalArgumentException("Unknown axis '" + axis.type().getLabel() + "' encountered.");
 		}
 	}
-	
+
 	static long[] reverseShape(long[] shape) {
 		long[] reversedShape = new long[shape.length];
 		for (int i = 0; i < shape.length; i++) {