TypeError: 'NoneType' object is not subscriptable

[manvirvirk]

File "D:\pytorch-nested-unet-master\dataset.py", line 63, in getitem
img_id + self.mask_ext), cv2.IMREAD_GRAYSCALE)[..., None])
TypeError: 'NoneType' object is not subscriptable

i m getting this error while training model, Can anyone help??

[4uiiurz1]

@manvirvirk
Does the file look like the following?

inputs
├── data-science-bowl-2018
|   ├── stage1_train
|   |   ├── 00ae65...
|   │   │   ├── images
|   │   │   │   └── 00ae65...
|   │   │   └── masks
|   │   │       └── 00ae65...            
|   │   ├── ...
|   |
|   ...
└── dsb2018_96
    ├── images
    └── masks
             └── 0
                     ├── 00ae65...
                     ...

[balag59]

Thank you for the implementation! I have the same issue with the latest commit. I'm training on an original dataset and this is how it looks:

Original_Dataset
images
1.jpg
masks
0
1.png

I'm am able to print all the img_ids successfully after reading them so the dataset is being read correctly I guess.

[4uiiurz1]

@balag59
Thanks for letting me know.
How do you set img_ext and mask_ext?

[manvirvirk]

i want to train on my own database. I have 20 colored images and respective masks. HOow to make directory of database. Thanks

[balag59]

@balag59
Thanks for letting me know.
How do you set img_ext and mask_ext?

I was able to rectify this error as I forgot to set the right mask dir in train.py. That error has been resolved now. Thank you for your time!

[manvirvirk]

did you train this model on your own database?? @balag59

[balag59]

did you train this model on your own database?? @balag59

Yes with just 1 class.

[manvirvirk]

Thanks for reply. I want to train it with my own database which contains 20 training images and respective masks. So what all changes we have to make. Please help as i m getting lot of errors. @balag59

[manvirvirk]

this is my directory:
input-dataset_name-image(it contains 20 images in png format)
masks(it contains 20 images in png format)
now what changes we need to make??? @balag59 @4uiiurz1

[balag59]

this is my directory:
input-dataset_name-image(it contains 20 images in png format)
masks(it contains 20 images in png format)
now what changes we need to make??? @balag59 @4uiiurz1

I didn't run into any issues as soon as I fixed my mask directory path. Make sure that you are setting the right path for the image and mask directories. According to what you've mentioned above, you aren't placing the masks class wise right? All your masks are directly under the masks directory? If so, then you will need to place it class wise as mentioned in the README starting with class 0.

[manvirvirk]

thanks . now i enetered in this error:
File "C:\Users\virkt\Anaconda3\envs\unet\lib\site-packages\sklearn\model_selection_split.py", line 1805, in _validate_shuffle_split
train_size)
ValueError: With n_samples=0, test_size=0.2 and train_size=None, the resulting train set will be empty. Adjust any of the aforementioned parameters.

[balag59]

thanks . now i enetered in this error:
File "C:\Users\virkt\Anaconda3\envs\unet\lib\site-packages\sklearn\model_selection_split.py", line 1805, in _validate_shuffle_split
train_size)
ValueError: With n_samples=0, test_size=0.2 and train_size=None, the resulting train set will be empty. Adjust any of the aforementioned parameters.

This error means that your image_ids directory is still not set correctly and that's why the n_samples is equal to 0. You will need to cross-check your image_ids path again and see if it's accurate.

[manvirvirk]

can you specify where in code i need to cross verify image_ids path??

[manvirvirk]

I m beginner so i m getting bit confused. Please specify where how to set image_ids path? Thanks for helping long way. @balag59

[manvirvirk]

this is how my directory looks like
input
dataset name
-images : contains 20 rgb images png format
-mask
-0 : contains 20 masks png format

[balag59]

can you specify where in code i need to cross verify image_ids path??

Check line 247

[manvirvirk]

in train.py file???

[4uiiurz1]

@manvirvirk Please try to change directory name input to inputs

[manvirvirk]

img_ids = glob(os.path.join('inputs', config['dataset'], 'images', '*' + config['img_ext']))
img_ids = [os.path.splitext(os.path.basename(p))[0] for p in img_ids]
i think these lines? what changes need to be made here? thanks @balag59

[manvirvirk]

oh got it

[manvirvirk]

its already inputs

[4uiiurz1]

ok...

[manvirvirk]

so what changes need to be made in image_ids path?

[4uiiurz1]

Please try the following code:

img_ids = glob(os.path.join('inputs', config['dataset'], 'images', '*' + config['img_ext']))
img_ids = [os.path.splitext(os.path.basename(p))[0] for p in img_ids]
print(img_ids)

[manvirvirk]

started training , thanks for helping long way :)

[manvirvirk]

i think print(img_ids) command worked:)

[manvirvirk]

did you do testing also???

[manvirvirk]

did you validate the model? @balag59

[manvirvirk]

how we get segmented image???

[4uiiurz1]

Please read README.