@Test
public void testInvalidDatasetTypesFail() throws AssertionError {
final Stream<DatasetType> allTypes = Arrays.stream(DatasetType.values());
allTypes
.filter(t -> t != DatasetType.BIT)
.forEach(
type -> {
dataset = datasetCreator.createDataset(type);
final boolean result = (boolean) ij.op().run(DatasetIsBinary.class, dataset);
final String typeClassName = dataset.getType().getClass().getName();
assertFalse("A Dataset of type " + typeClassName + " should not be binary", result);
});
}
public void dummySegmentation() {
List<Dataset> predictDataset = new ArrayList<>(imgDatasets.size());
INDArray predict = dummyClassification();
int currIdx = 0;
for (int ds = 0; ds < imgDatasets.size(); ds++) {
List<List<int[]>> profilesSet = testData.get(ds).getProfiles();
predictDataset.add(imagePlaneService.createEmptyPlaneDataset(imgDatasets.get(ds)));
Dataset currDataset = predictDataset.get(ds);
RandomAccess<RealType<?>> randomAccess = currDataset.randomAccess();
for (List<int[]> p : profilesSet) {
for (int i = 0; i < p.size(); i++) {
randomAccess.setPosition(p.get(i)[0], 0);
randomAccess.setPosition(p.get(i)[1], 1);
double value = predict.getRow(currIdx).getDouble(i);
if (value != 0) randomAccess.get().setReal(value);
}
}
Overlay[] overlays = BinaryToOverlay.transform(context, currDataset, true);
// List<Overlay> overlayList = Arrays.asList(BinaryToOverlay.transform(context,
// currDataset, true));
overlayStatService.setRandomColor(Arrays.asList(overlays));
ImageDisplay display = new DefaultImageDisplay();
context.inject(display);
display.display(currDataset);
for (Overlay o : overlays) {
display.display(o);
}
uis.show(display);
}
}
/**
* Generate the corresponding sequences of labels (0, 1) for a set of profiles, depending on if
* the pixel belongs to the membrane or not
*
* @param trainingSet Set of intensity profiles extracted from an image, for the training of the
* neural network and for that we need to label.
* @param labeledDs Dataset where pixels belonging to membranes have been drawn.
*/
public void generateConfirmationSet(ProfilesSet trainingSet, Dataset labeledDs) {
RandomAccess<RealType<?>> randomAccess = labeledDs.randomAccess();
List<int[]> labeledProfiles = new ArrayList<>(trainingSet.getProfiles().size());
for (int i = 0; i < trainingSet.getProfiles().size(); i++) {
List<int[]> profile = trainingSet.getProfiles().get(i);
int[] labels = new int[trainingSet.getMaxLenght()];
Arrays.fill(labels, 0);
for (int j = 0; j < profile.size(); j++) {
int[] point = profile.get(j);
randomAccess.setPosition(point[0], 0);
randomAccess.setPosition(point[1], 1);
double pixelValue = randomAccess.get().getRealDouble();
if (pixelValue == COLOR) {
for (int k = j - membraneWidth.getValue() / 2;
k <= j + membraneWidth.getValue() / 2;
k++) {
if (k < 0 || k > labels.length) continue;
else labels[k] = 1;
}
break;
}
labeledProfiles.add(labels);
}
confirmationSet.add(labeledProfiles);
}
}
@SuppressWarnings({"unchecked", "rawtypes"})
private RealImageFunction<?, DoubleType> imgFunc(final Dataset ds) {
final Img<? extends RealType<?>> imgPlus = ds.getImgPlus();
return new RealImageFunction(imgPlus, new DoubleType());
}