Unsupported type: <java class 'net.imglib2.type.numeric.real.DoubleType'> in ij.py.from_java()

[YunxiaoWangwww]

Hi, I am running the Gray level co-occurrence matrices example in pyimagej, but it seems ij.op().haralick().correlation() and ij.op().haralick().differenceVariance() output java class net.imglib2.type.numeric.real.DoubleType, which cannot be converted into Python using ij.py.from_java().

import matplotlib.pyplot as plt
import pandas as pd
from matplotlib import patches
MatrixOrientation2D = imagej.sj.jimport('net.imagej.ops.image.cooccurrenceMatrix.MatrixOrientation2D')
orientations = [
    MatrixOrientation2D.ANTIDIAGONAL,
    MatrixOrientation2D.DIAGONAL,
    MatrixOrientation2D.HORIZONTAL,
    MatrixOrientation2D.VERTICAL
]
    corr = ij.op().haralick().correlation(img, gray_levels, dist, angle)
    diff = ij.op().haralick().differenceVariance(img, gray_levels, dist, angle)
    corr = ij.py.from_java(corr)
    diff = ij.py.from_java(diff)
    return (corr.value, diff.value)
def process_crops(crops, gray_levels: int, dist: int) -> pd.DataFrame:
    glcm_mean_results = []
    for key, value in crops.items():
        glcm_angle_results = []
        image = ij.py.to_dataset(value) # convert the view to a net.imagej.Dataset
            glcm_angle_results.append(run_glcm(image, gray_levels, dist, angle))
        corr_mean = sum(x[0] for x in glcm_angle_results) / len(glcm_angle_results)
        diff_mean = sum(x[1] for x in glcm_angle_results) / len(glcm_angle_results)
        glcm_mean_results.append((corr_mean, diff_mean))
    return pd.DataFrame(glcm_mean_results, columns=['corr', 'diff'])`
gray_levels = 128 # a value lower than the bit depth of the image and typically a power of 2
dist = 7 # distance in pixels# set up GLCM parameters
gray_levels = 128 # a value lower than the bit depth of the image and typically a power of 2
dist = 7 # distance in pixels
data = ij.io().open('https://media.imagej.net/workshops/data/2d/hela_hiv_gag-yfp.tif')
data_xarr = ij.py.from_java(data)
ij.py.show(data_xarr * 12, cmap='Greys_r') # multiply by 12 to better visualize the data (doesn't change source)
crops = {
    "cyto1": data[318: 368, 369: 419], # cell 1 cytoplasm crop
    "cyto2": data[130: 180, 355: 405], # cell 2 cytoplasm crop
    "cyto3": data[87: 137, 194: 244],  # cell 3 cytoplasm crop
    "cyto4": data[256: 306, 43: 93],   # cell 4 cytoplasm crop
    "bkg1": data[19: 69, 57: 107],     # background 1 crop
    "bkg2": data[263: 313, 221: 271]   # background 2 crop
}
crop_coords = {
    "cyto1": (318, 369),
    "cyto2": (130, 355),
    "cyto3": (87, 194),
    "cyto4": (256, 43),
    "bkg1": (19, 57),
    "bkg2": (263, 221)
}
plt.style.use('ggplot')
df = process_crops(crops, gray_levels, dist)
df["name"] = ["cyto1", "cyto2", "cyto3", "cyto4", "bkg1", "bkg2"]
plt.scatter(df['corr'], df['diff'])
for i in range(len(df)):
    plt.annotate(f"{df['name'][i]}", (df['corr'][i], df['diff'][i]))
plt.xlabel('corr')
plt.ylabel('diff')
plt.title('GLCM texutre plot')
plt.show()


with the error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[10], line 5
      2 plt.style.use('ggplot')
      4 # process the dict of crops and add crop names to the output dataframe
----> 5 df = process_crops(crops, gray_levels, dist)
      6 df["name"] = ["cyto1", "cyto2", "cyto3", "cyto4", "bkg1", "bkg2"]
      8 # plot the data in a matplotlib scatter plot
Cell In[5], line 36, in process_crops(crops, gray_levels, dist)
     34 # compute the correlation and difference variance textures at all orientations
     35 for angle in orientations:
---> 36     glcm_angle_results.append(run_glcm(image, gray_levels, dist, angle))
     37 # calculate the mean of the angle results
     38 corr_mean = sum(x[0] for x in glcm_angle_results) / len(glcm_angle_results)
Cell In[5], line 16, in run_glcm(img, gray_levels, dist, angle)
     13 diff = ij.op().haralick().differenceVariance(img, gray_levels, dist, angle)
     15 # convert to Python float
---> 16 corr = ij.py.from_java(corr)
     17 diff = ij.py.from_java(diff)
     19 return (corr.value, diff.value)
File ~/.pyenv/versions/3.8.19/lib/python3.8/site-packages/imagej/__init__.py:288, in ImageJPython.from_java(self, data)
    279 """Convert supported Java data into Python equivalents.
    280 
    281 Converts Java objects (e.g. 'net.imagej.Dataset') into the Python
   (...)
    285 :return: A Python object converted from Java.
    286 """
    287 # todo: convert a dataset to xarray
--> 288 return sj.to_python(data)
File ~/.pyenv/versions/3.8.19/lib/python3.8/site-packages/scyjava/__init__.py:700, in to_python(data, gentle)
    698 except TypeError as exc:
    699     if gentle: return data
--> 700     raise exc
File ~/.pyenv/versions/3.8.19/lib/python3.8/site-packages/scyjava/__init__.py:697, in to_python(data, gentle)
    695 start_jvm()
    696 try:
--> 697     return _convert(data, py_converters)
    698 except TypeError as exc:
    699     if gentle: return data
File ~/.pyenv/versions/3.8.19/lib/python3.8/site-packages/scyjava/__init__.py:266, in _convert(obj, converters)
    264 suitable_converters = filter(lambda c: c.predicate(obj), converters)
    265 prioritized = max(suitable_converters, key = lambda c: c.priority)
--> 266 return prioritized.converter(obj)
File ~/.pyenv/versions/3.8.19/lib/python3.8/site-packages/scyjava/__init__.py:345, in _raise_type_exception(obj)
    344 def _raise_type_exception(obj: Any):
--> 345     raise TypeError('Unsupported type: ' + str(type(obj)))
TypeError: Unsupported type: <java class 'net.imglib2.type.numeric.real.DoubleType'>
``

[YunxiaoWangwww]

I used net.imglib2.type.numeric.real.DoubleType.getRealFloat to convert it into a Java Float, but the number is not exactly the same as before.

[elevans]

Hi @YunxiaoWangwww, what version of PyImageJ are you using? I just ran the notebook without errors with the latest release version 1.5.0 and openjdk 11.0.23. Also what version of scyjava do you have installed in your environment?

[ctrueden]

I used net.imglib2.type.numeric.real.DoubleType.getRealFloat to convert it into a Java Float, but the number is not exactly the same as before.

You can convert it from DoubleType to Java Double using getRealDouble(), which will be lossless in precision.