At the moment, a single ROI can be stored in a ZIP file, or the call can be used multiple times to append ROIs into a ZIP file, which is however much slower than directly using zipfile to create a ZIP with multiple ROIs in one go. As a feature suggestion, a function to create a ZIP file with a list of ROIs would be very nice.

Hello Christoph,

Hope you're doing well!

I've installed roifile and wanted to get familiar with its code and features by running the roifile_demo.py. It looks like the code is trying to open a missing demo file, i.e. the file box.tif.

Is the tests folder missing or part of another project?'

Best regards,
Chris

C:\Users\User\Desktop\roifile-master>roifile_demo.py
Traceback (most recent call last):
  File "C:\Users\User\Desktop\roifile-master\roifile_demo.py", line 31, in <module>
    with TiffFile('tests/box.tif') as tif:
  File "C:\python\lib\site-packages\tifffile\tifffile.py", line 2977, in __init__
    fh = FileHandle(arg, mode=mode, name=name, offset=offset, size=size)
  File "C:\python\lib\site-packages\tifffile\tifffile.py", line 9446, in __init__
    self.open()
  File "C:\python\lib\site-packages\tifffile\tifffile.py", line 9459, in open
    self._fh = open(self._file, self._mode)
FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\User\\Desktop\\roifile-master\\tests\\box.tif'

Thanks @cgohlke for this great library!

It seems like roiwrite burns the ROIs iterable to write if that iterable is a generator-like object / traversable-once, resulting in an empty output file. Would you consider changing implementation to accommodate generator as the argument for ROIs?

Hi,

I am loading .roi files from a .zip and modifying them using ImageJRoi.fromfile(). I have noticed that in doing so it is reindexing the files so if you load the modified .zip folder into ImageJ the labels (1-n) are different than the original. It essentially sorts it in the order the .roi files are loaded.

Is there a value within the ImageJRoi object that can be used to identify the files original index value and modified in the new object to maintain the same index?

I have a large image (36000 x 45000) pixels and it's basically a mask of multiple objects. Since the image is big, some of the objects have coordinates bigger than byte mode defined in struct.pack. I am a uploading the ".npy" array to reproduce the error.

x = 'provided numpy array' roi = ImageJRoi.frompoints(x) roi.tofile()

I hope you can help, because it works for only the objects with max coordinates < 32767.

Thank you

coords.zip

Dear,

Thanks for this python package, is there an example on how to write PointRois?

Best
Benjamin

Hey,

I've had the necessity of correcting roi coordinates and I did it like this:

    # Replace roi coordinates by new ones 
    int_diff = np.array([*map(min, new_coordinates.T)])[:,None] 
    roi1.__dict__['integer_coordinates']  = (new_coordinates.T - int_diff ).round().astype(int)
    roi1.__dict__['subpixel_coordinates'] = new_coordinates.T.round().astype(float) 

When I print the roi1.coordinates(), I get the new_coordinates as I intend

Cell408_3717_1346
[[3773. 1367.]
[3785. 1348.]
[3763. 1315.]
[3743. 1303.]
[3733. 1298.]
[3720. 1294.]
[3699. 1312.]
[3671. 1346.]
[3668. 1352.]
[3659. 1360.]
[3661. 1370.]
[3659. 1375.]
[3662. 1380.]
[3664. 1381.]
[3673. 1389.]
[3682. 1391.]
[3687. 1388.]
[3690. 1389.]
[3711. 1390.]
[3754. 1374.]
[3762. 1373.]
[3764. 1370.]
[3773. 1367.]]

However, when I save the roifile ( with roifile.roiwrite('roifile.zip', list_several_output_rois)) and then read that roifile.zip file in python or load it into imageJ, the roi.coordinates() result of the exact same roi looks like this:

Cell408_3717_1346
[[3773. 3682.]
[3785. 3687.]
[3763. 3690.]
[3743. 3711.]
[3733. 3754.]
[3720. 3762.]
[3699. 3764.]
[3671. 3773.]
[3668. 1367.]
[3659. 1348.]
[3661. 1315.]
[3659. 1303.]
[3662. 1298.]
[3664. 1294.]
[3673. 1312.]]

It colapses the points and moves some of the missing points x coordinates into the y coordinate of previous points. Can you help me solve this issue? Is it a bug of roifile.roiwrite() ?

Kind regards,
Ricardo

ImagejRoi.frompoints picks a random index to create a (I guess supposedly) unique name per ROI.
However it occurred multiple times for me in some batch processing jobs with many thousand ROIs, that collisions occurred.
I'm well aware that there is no easy solution to this.
Potential fixes would be to raise the bits of the random number to make a collision more unlinkely, or to have a singleton variable in the module as a ROI counter (still, e.g. ROIs arriving from multiple processes could have colliding IDs).
Do you think there is a good solution to this? I guess I will workaround this for now by providing an incremented index by my code.

For example,

import numpy
numpy.random.seed(0)

def collide():
    set_ = set()
    while True:
        ri = numpy.random.randint(0, 2**31 - 1)
        if ri in set_:
            return len(set_)
        else:
            set_.add(ri)

for _ in range(5): print(collide())

Yields some not-so-super-high collission counts:

67205
64247
60800
105507
27671