search

BAMresearch / automatic-sem-image-segmentation

Workflow for Simulation and Automatic Semantic Segmentation of Electron Microscopy Images
Other
43 stars 11 forks source link

Tuple index out of range #1

Open sp4m4r opened 3 years ago

sp4m4r commented 3 years ago

Ladies and Gentlemen,

thank you for sharing this very interesting work. I'm not a specialist in neural networks but anyway very interested in trying to apply your method.

I updated and installed all required packeges, using pip(3) and python 3.9.5. However, using the example from the folder "Archive/Automatic_SEM_Image_Segmentation' I'm getting the error:

File "C:\Users***\Desktop\automatic-sem-image-segmentation-master\Archive\Automatic_SEM_Image_Segmentation\startProcess.py", line 347, in imgTiles = imgTiles.reshape([imgTiles.shape[0], imgTiles.shape[1], imgTiles.shape[2], 1]) / 255.0 IndexError: tuple index out of range

As the template is used right as is I can not figure out what's causing the problem here. Any suggestion is very welcome, thank you in advance!

iMasak commented 2 years ago

I see the same issue while testing the archive file and data. Are there any updates or solutions? thank you,

bruehle commented 2 years ago

This error is most probably due to changes in one of the dependencies that is not backwards compatible. I tested it with the following dependency versions, and it ran without a problem (tested on Windows 10, python 3.7.4, CUDA 11.1, cuDNN 7.6):

numpy==1.18.5 tensorflow==2.2.0 tensorflow_addons==0.11.2 imageio==2.5.0 scikit-image==0.15.0 Pillow==8.2.0 scipy==1.4.1 opencv-python==4.5.2.54 matplotlib==3.5.2 tqdm==4.62.3 opensimplex==0.3

I hope I will eventually get around uploading an updated and refactored version of the code that will also run with more recent versions of the dependencies, but in the meantime, please let me know if using the above-mentioned versions fixed your issue.

PS: If you are getting an error during saving the CycleGAN model, you might also need to shorten the path name (the default after downloading, i.e., C:\users\\Documents\automatic-sem-image-segmentation-master\Automatic_SEM_Image_Segmentation might create too long pathnames after appending the model save path, C:\users\\Documents\segmentation should however work).

pramethg commented 2 years ago

Ladies and Gentlemen,

thank you for sharing this very interesting work. I'm not a specialist in neural networks but anyway very interested in trying to apply your method.

I updated and installed all required packeges, using pip(3) and python 3.9.5. However, using the example from the folder "Archive/Automatic_SEM_Image_Segmentation' I'm getting the error:

File "C:\Users***\Desktop\automatic-sem-image-segmentation-master\Archive\Automatic_SEM_Image_Segmentation\startProcess.py", line 347, in imgTiles = imgTiles.reshape([imgTiles.shape[0], imgTiles.shape[1], imgTiles.shape[2], 1]) / 255.0 IndexError: tuple index out of range

As the template is used right as is I can not figure out what's causing the problem here. Any suggestion is very welcome, thank you in advance!

Were you able to fix this?

bruehle commented 2 years ago

The error should not occur when using the versions mentioned in my comment above. Please use

pip install -r requirements.txt

to install the packages with the versions mentioned above. And please also make sure you also have the correct versions of python, CUDA, and cuDNN installed (if I remember correctly, tensorflow 2.2 is not compatible with python 3.9).

I am currently working on an updated and refactored version of the code that will run with newer versions of python and tensorflow which I hope to release next week, but in the meantime, please try with the versions mentioned above.