division by zero

[DrraBL]

hi I had the following error "" sum(train_loss)/len(train_loss), sum(val_loss)/len(val_loss), time.time()-start_time)) "" even i put 40 samples for train-data and 10 samples for val-data, I think that len(train_loss) and len(val_loss) have zero value. Please let me know if you have any idea.

[amanbasu]

Looks like your data is not getting read properly. Make sure the file path is valid.

To better debug, you can try printing the stacktrace inside the except block.

[DrraBL]

hi thank you for your reply I solved this issue and i had another one about the nu;ber of slices as said the below error:

Traceback (most recent call last): File "train.py", line 361, in train() File "train.py", line 330, in train loss, _, summary = sess.run([cost, opt, merged], feed_dict={is_training: True}) File "/home/adel/anaconda3/envs/PCS_PhD/lib/python3.7/site-packages/tensorflow/python/client/session.py", line 950, in run run_metadata_ptr) File "/home/adel/anaconda3/envs/PCS_PhD/lib/python3.7/site-packages/tensorflow/python/client/session.py", line 1173, in _run feed_dict_tensor, options, run_metadata) File "/home/adel/anaconda3/envs/PCS_PhD/lib/python3.7/site-packages/tensorflow/python/client/session.py", line 1350, in _do_run run_metadata) File "/home/adel/anaconda3/envs/PCS_PhD/lib/python3.7/site-packages/tensorflow/python/client/session.py", line 1370, in _do_call raise type(e)(node_def, op, message) tensorflow.python.framework.errors_impl.InvalidArgumentError: Cannot batch tensors with different shapes in component 0. First element had shape [512,512,46] and element 1 had shape [512,512,23]. [[node IteratorGetNext (defined at train.py:53) ]]

Errors may have originated from an input operation. Input Source operations connected to node IteratorGetNext: IteratorV2 (defined at train.py:49)

Original stack trace for 'IteratorGetNext': File "train.py", line 53, in next_item = iterator.get_next()

---

each files have it own number of slices so how i can reshape all the files with the same shape?? thank you

[amanbasu]

Please use resizing.py to pre-process your data.

[DrraBL]

thank you for your reply . i used the script resizing.py to resize the input data, but no result, I mean the script compiled successfully but the file 's shape don't change and i got the same error when i run train.py the code of the resizing file is below:

---

''' The code for resizing has been taken from https://gist.github.com/zivy/79d7ee0490faee1156c1277a78e4a4c4 ''' import os import glob import numpy as np import SimpleITK as sitk

def resample(img, new_size, interpolator): dimension = img.GetDimension()

# Physical image size corresponds to the largest physical size in the training set, or any other arbitrary size.
reference_physical_size = np.zeros(dimension)

reference_physical_size[:] = [(sz-1)*spc if sz*spc>mx  else mx for sz,spc,mx in zip(img.GetSize(), img.GetSpacing(),     reference_physical_size)]

# Create the reference image with a zero origin, identity direction cosine matrix and dimension     
reference_origin = np.zeros(dimension)
reference_direction = np.identity(dimension).flatten()
reference_size = new_size
reference_spacing = [ phys_sz/(sz-1) for sz,phys_sz in zip(reference_size, reference_physical_size) ]

reference_image = sitk.Image(reference_size, img.GetPixelIDValue())
reference_image.SetOrigin(reference_origin)
reference_image.SetSpacing(reference_spacing)
reference_image.SetDirection(reference_direction)

# Always use the TransformContinuousIndexToPhysicalPoint to compute an indexed point's physical coordinates as 
# this takes into account size, spacing and direction cosines. For the vast majority of images the direction 
# cosines are the identity matrix, but when this isn't the case simply multiplying the central index by the 
# spacing will not yield the correct coordinates resulting in a long debugging session. 
reference_center = np.array(reference_image.TransformContinuousIndexToPhysicalPoint(np.array(reference_image.GetSize())/2.0))

# Transform which maps from the reference_image to the current img with the translation mapping the image
# origins to each other.
transform = sitk.AffineTransform(dimension)
transform.SetMatrix(img.GetDirection())
transform.SetTranslation(np.array(img.GetOrigin()) - reference_origin)
# Modify the transformation to align the centers of the original and reference image instead of their origins.
centering_transform = sitk.TranslationTransform(dimension)
img_center = np.array(img.TransformContinuousIndexToPhysicalPoint(np.array(img.GetSize())/2.0))
centering_transform.SetOffset(np.array(transform.GetInverse().TransformPoint(img_center) - reference_center))
centered_transform = sitk.Transform(transform)
centered_transform.AddTransform(centering_transform)
# Using the linear interpolator as these are intensity images, if there is a need to resample a ground truth 
# segmentation then the segmentation image should be resampled using the NearestNeighbor interpolator so that 
# no new labels are introduced.

return sitk.Resample(img, reference_image, centered_transform, interpolator, 0.0)

new_size = [144,144,50]
interp = sitk.sitkNearestNeighbor # for labels interp = sitk.sitkLinear # for input features

for file in sorted(glob.glob('../train-data/Case*_segmentation.mhd')):

uncomment when resizing input images

file = file.replace('_segmentation', '')
img = sitk.ReadImage(file)
reshaped = resample(img, new_size, interp)
sitk.WriteImage(reshaped, file)
print(file, end='\r')

---

please let me know and thank in advance

[dragon-king-creater]

Hello, I have trained and saved the parameters, but I want to know whether the final result of your prediction program is an array? Why not a distorted image? And why my prediction result is an empty array. I hope to get your reply

[amanbasu]

size the input data, but no result, I mean the script compiled successfully but the file 's shape don't change and i got the same error...

Please check whether the path you are using is correct or not. Try to debug the code. Also, keep in mind to use interp = sitk.sitkLinear and file = file.replace('_segmentation', '') for inputs and interp = sitk.sitkNearestNeighbour for labels (as mentioned in the file, check comments).

[amanbasu]

Hello, I have trained and saved the parameters, but I want to know whether the final result of your prediction program is an array? Why not a distorted image? And why my prediction result is an empty array. I hope to get your reply

The output of the file predict.py is a list of arrays (predictions). The list might be empty if the file path is incorrect i.e. no image is getting read. Also, what do you mean by distorted image?

[DrraBL]

Hi @amanbasu thank you for your previous replies so I trained the model and saved the weight, but i don't know if it's done successfully or not. now i am trying to run predict file and i got the following error: Traceback (most recent call last): File "predict.py", line 266, in predictions = predict(init_epoch) File "predict.py", line 252, in predict saver = tf.train.import_meta_graph('/result/weightsepoch{0}.ckpt.meta'.format(init_epoch)) raise ValueError(str(e)) ValueError: Cannot add function '__inference_Datasetmap_12' because a different function with the same name already exists.

Please if you have any idea about this error let me know otherwise I wanna ask you about the testing stage because i want to measure the dice coefficient for the test dataset? I am looking forward to hear from you.

[amanbasu]

For the error, you can try restarting the tensorflow session. Perhaps that might help.

[DrraBL]

HI @amanbasu thank you for your previous replies, I resolved the previous error by changing the tf version, i upgraded it but i have another problem about the compatibility, i tried to do the necessary modification like: import tensorflow.compat.v1 as tf tf.disable_v2_behavior() and about the predict.py file , I changed the following line: iterator = tf.data.Iterator.from_structure(valDataset.output_types, valDataset.output_shapes) to iterator = tf.data.Iterator.from_structure(tf.compat.v1.data.get_output_types(valDataset), tf.compat.v1.data.get_output_types(valDataset)) but I meet this error: TypeError: Dimension value must be integer or None or have an index method, got value 'tf.float32' with type '<class 'tensorflow.python.framework.dtypes.DType'>' I coudn't resolve it, please let me know.

[amanbasu]

iterator = tf.data.Iterator.from_structure(tf.compat.v1.data.get_output_types(valDataset), tf.compat.v1.data.get_output_types(valDataset))

You are using data.get_output_types for both arguments. Perhaps that is the issue.

[DrraBL]

HI @amanbasu Thank you for your previous replies, the training is done but for the testing i did not get any results, however my goals to get segmented images and calculate the Dice coefficient. sorry if I bothered with my questions but i new in this domain. Looking forward hearing from you

kindly,

[amanbasu]

Try to look into the predict.py file. On line 265, the predictions are returned as a list. You need to save the results here.