在与Jupyter Notebook不同的环境中运行python脚本

Question

I have a main Jupyter Notebook file that operates on a Slicer kernel. It opens slicer and makes the user save the labels. I know what to run a deep learning model in PyTorch on a Python 3 kernel. To do this, I first need to transform the labels to the format that I need which is tiff.

I managed to do it in Spyder.

import numpy as np
from imageio import imwrite
import os

## define where the masks are saved as a .npy
labelpath="../temp/label.npy" # changed for privacy
## define path where the new masks should be saved as .tiff files
labelpathsave="../temp/" # changed for privacy

## open the object
label=np.load(labelpath)

number_files=np.shape(label)[0]

for i in range(0, number_files):
   string="label"+str(i+1)+".tiff"
   labelpath=os.path.join(labelpathsave, string)
   currentlabel=label[i]
   imwrite(labelpath, currentlabel)

This above works, but trying to execute this python file in the Jupyter Notebook.

execfile('../scripts/Maja/transform_label.py') # changed for privacy

I get "ImportError: No module named imageio".

How can I make it work execute "transform_label.py" given that it runs in a different environment?

(I need to do it in separate files because I cannot get the PyTorch package to work on a Slicer kernel)

Answer 1

installing imageio

for python 2

pip install imageio --user

for python 3

pip3 install imageio --user

For Conda Environments

conda install -c conda-forge imageio

NOTE : Must check first which version of python is being running on Notebook.

After installation ( The installation includes all necessary lib that are needed to import .) Rerun the above code!!!...

Answer 2

I think that the script is not working because the imageio module cannot be found. Have you tried to save imageio in a directory where you know that other modules can be loaded by Jupyter notebooks? Maybe it is just a matter of the directories, where jupyter notebooks looks for modules.

Or it could also be that for spyder (if you use it through the anaconda distribution) had this imageio package preinstalled. So you could also try and just check in jupyter if you have the package installed there.