Welcome to Cervical Fuzzy Ensemble Discussions!

[rishavpramanik]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[sangin-fatemeh]

I have downloaded 5 compressed files as follows:

• im_Dyskeratotic.7z
• im_Koilocytotic.7z
• im_Superficial-Intermediate.7z
• im_Metaplastic.7z
• im_Parabasal.7z
  Now, could you please explain me how to divide these into the required foldering format, which is:
  +-- data
  | +-- .
  | +-- train
  | | +--class A
  | | +--class B
  | | ...
  | +-- val
  | | +--class A
  | | +--class B

Thank you a lot in advance.

[rishavpramanik]

Hi Fatemeh,
Thanks for your query, you may refer to this discussion on stack overflow https://stackoverflow.com/questions/53074712/how-to-split-folder-of-images-into-test-training-validation-sets-with-stratified

Here 2-3 approaches are discussed in detail. Hope this helps!
Also, do not forget to unzip before you split them

[yodhcn]

Q1: Can you describe how to resize the training samples? I couldn't find the relevant code.

• Forced stretch?
• Center crop first and then scale?
• Scale first and then fill with 0?

Q2: Why scale the image to 256256 instead of 224224, doesn't this behavior affect the pretrained weights?

Q3: Is the test set not splited in the experiment?

Thank you very much.

[rishavpramanik]

Thanks @yodhcn for getting in touch,
A1: Actually, the dataset itself contains images all resized with (256,256). So we didn't have to apply any resizing operations.
A2: We did not scale it to 25256, it does not affect to a large extent the pretrained weights. You may refer to this discussion: https://pyimagesearch.com/2019/06/24/change-input-shape-dimensions-for-fine-tuning-with-keras/
A3: Since we are doing 5-fold CV, we don't have an explicit test set. We followed the experimental protocols of the paper https://www.nature.com/articles/s41598-021-93783-8

I hope I was able to answer your queries. Thanks once again for your question

[yodhcn]

I have downloaded 5 compressed files as follows:
https://www.cs.uoi.gr/~marina/sipakmed.html

• im_Dyskeratotic.7z
• im_Koilocytotic.7z
• im_Superficial-Intermediate.7z
• im_Metaplastic.7z
• im_Parabasal.7z

But the pictures in folder ./CROPPED have different aspect ratios.
E.g im_Dyskeratotic/CROPPED/002_01.bmp (127 * 106)

[yodhcn]

Q4: Should I use one-hot encoding here?
https://github.com/rishavpramanik/CervicalFuzzyDistanceEnsemble/blob/main/utils/k_fold_separate.py#L55
https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator#examples

y_train = tf.keras.utils.to_categorical(encode_y(y_train))

Thank you for your patience.

[rishavpramanik]

You can use any one of them, depending on what is your loss, for example if you use sparse CCE. This will throw an error. Whereas MSE will require this step

[vaziri-ai]

hi. thank you for the codes. I'm running the code but i get this error. can you pls help me?
Error:

thanks in advance