E ValueError: numpy.ufunc size changed, may indicate binary incompatibility. Expected 216 from C header, got 192 from PyObject about cosmos-tools HOT 22 OPEN

What version of numpy are you using? I think these errors have to do with a versioning issue.

Can you try using an older version and see if that helps?
e.g.: pip install numpy==1.16.1

from cosmos-tools.

I used the file 'cosmos3requirements.yml' and it contains numpy version 1.14.0
I'd tried installing numpy 1.16.1 in the environment and got this error

ERROR: After October 2020 you may experience errors when installing or updating packages. This is because pip will change the way that it resolves dependency conflicts.

We recommend you use --use-feature=2020-resolver to test your packages with the new resolver before it becomes the default.

tensorflow 1.10.0 requires numpy<=1.14.5,>=1.13.3, but you'll have numpy 1.16.1 which is incompatible.
statsmodels 0.12.2 requires pandas>=0.21, but you'll have pandas 0.20.3 which is incompatible.
moviepy 0.2.3.2 requires decorator==4.0.11, but you'll have decorator 5.1.1 which is incompatible.

Should I install tensorflow, statsmodel, and moviepy with another version?

from cosmos-tools.

Try:

pip install numpy==1.14.0
pip install pandas==0.20.3
pip install statsmodels==0.10.2

from cosmos-tools.

I think the issue is that we didn't specify statsmodels version

from cosmos-tools.

The version conflict problem had been solved but I encountered this error.

I checked the directory and the files exist like below.

from cosmos-tools.

This is an interesting issue.

The path ...\top\top_out.mat\keep_cells.npz doesn't make sense.

Looking in the function that's called in cosmos_dataset.py, maybe line 105 needs to be changed as follows:

vid_dir = vid_path.rsplit("/", 1)[0]
to
vid_dir = vid_path.rsplit(os.path.sep, 1)[0]

Can you tell us if that solves the problem?

from cosmos-tools.

I tried changing all of vid_path.rsplit("/", 1)[0] in line 105 and other lines.
It looks working, but I meet warning message like below.
Can I ignore this warning or have to fix something else?

from cosmos-tools.

I don't get this warning when I run the tests on my end (I just now successfully installed on a new computer macOS 10.15.7)

Is it possible for you to determine what the invalid value is, i.e. by adding print statements or a debug breakpoint (i.e. add flag --pdb when you call pytest, like here: https://stackoverflow.com/questions/2678792/can-i-debug-with-python-debugger-when-using-py-test-somehow) ? Nans should have been removed in the lines immediately preceding line 570.

from cosmos-tools.

I get this warning with anaconda prompt at windows.
And all the elements in mean_block['top_focus'] and mean_block['bot_focus'] was 'NaN' when I debugged and printed those variables.
But when I use Ubuntu, I does successfully work.

from cosmos-tools.

Interesting...we never ran it on Windows, so it's not unreasonable to me that there might be an issue (i.e. with file paths).

from cosmos-tools.

I have a few more questions.
In order to try this software, I used the script 'import_raw_cosmos_data.py' and encountered some difficulty.

When I imported the example data provided with this software and select ROIs, I got this error.

So I downloaded the raw data video file from you paper, and then converted its file format from mp4 to ome.tif which has 300 frames using ImageJ plug-in 'Bioformat'. And When I tried this converted file, I got this new error.

I wonder whether I missed something important.

Additionally, when I select ROIs, do the 'top_ROI' and 'bot_ROI' mean central area focused image and that of lateral area respectively? Then should I select identical area as top and bot ROIs if I try this software with conventional macroscope?

from cosmos-tools.

from cosmos-tools.

Oh, I see. Thank you for your answers.
Still I get the same error when I try it with smaller ROI.

What video format is when you get raw data files?
Is the raw data saved as ome.tif format directly when you get the raw data, or saved as other format and converted to ome.tif?

from cosmos-tools.

from cosmos-tools.

from cosmos-tools.

Yes, I'd just tried choosing cranial-window-size ROI, and got same error.
Is it possible to get this error because of small number of frames of the example raw data file?
(It contains just 6 frames and I get Index Error as the captured Image shows)

from cosmos-tools.

from cosmos-tools.

I'll try it and check if it works well.
Thank you so much.

from cosmos-tools.

Duplicating the image stacks solved the index error.

The matlab window show me like below when I run CNMF-E through 'import_raw_cosmos_data.py',

Does anything exist I should change?

from cosmos-tools.

from cosmos-tools.

I added matlab path and it make it work.
However, the saving error occurs when running cnmfe.
How can I solve this problem?

from cosmos-tools.

from cosmos-tools.