search

IAmSuyogJadhav / 3d-mri-brain-tumor-segmentation-using-autoencoder-regularization

Keras implementation of the paper "3D MRI brain tumor segmentation using autoencoder regularization" by Myronenko A. (https://arxiv.org/abs/1810.11654).
MIT License
369 stars 114 forks source link

Zero Dice Coefficient #33

Closed vijayannepu closed 4 years ago

vijayannepu commented 4 years ago

Iam trying this approach on 2019 brats dataset with same model. The VAE dice coefficient (reconstructed image ) is around 0.90 but the dice coefficient for the segmentation label is 0.0000e+00.

Here is the model.fit method and output. Output is same for even 10 to 30 epochs.

model.fit(data, [labels, data],validation_split=0.1, batch_size=1,epochs=1,verbose=1,callbacks=[es,mc])

Train on 301 samples, validate on 34 samples Epoch 1/1 301/301 [==============================] - 6902s 23s/step - loss: 0.0413 - Dec_GT_Output_loss: -0.0065 - Dec_VAE_Output_loss: 0.0478 - Dec_GT_Output_dice_coefficient: 0.0065 - Dec_VAE_Output_dice_coefficient: 0.7997 - val_loss: 0.0229 - val_Dec_GT_Output_loss: 0.0000e+00 - val_Dec_VAE_Output_loss: 0.0229 - val_Dec_GT_Output_dice_coefficient: 0.0000e+00 - val_Dec_VAE_Output_dice_coefficient: 0.9005

Can some one help me..?

675492062 commented 4 years ago

I have trained for 50 epochs on brats2018. The final results as below loss: 0.0402 - Dec_GT_Output_loss: -0.0011 - Dec_VAE_Output_loss: 0.0412 - Dec_GT_Output_dice_coefficient: 0.0011 - Dec_VAE_Output_dice_coefficient: 0.8081

Crispy13 commented 4 years ago

In my experience, GT dice coefficient(train set) became acceptable (0.7~) after training for more than 200 epochs. I used only 2 channels (T1CE, T2) of the data, so the result may be different from that.

vijayannepu commented 4 years ago

No I am still in search of solution. I don't know why they have closed the thread.Even for 200 epochs it did not worked. I have tried it on 2018 dataset even. Please help if you were able to sort it out.

On Fri, 14 Feb 2020 at 16:03, sneh-debug notifications@github.com wrote:

Iam trying this approach on 2019 brats dataset with same model. The VAE dice coefficient (reconstructed image ) is around 0.90 but the dice coefficient for the segmentation label is 0.0000e+00.

Here is the model.fit method and output. Output is same for even 10 to 30 epochs.

model.fit(data, [labels, data],validation_split=0.1, batch_size=1,epochs=1,verbose=1,callbacks=[es,mc])

Train on 301 samples, validate on 34 samples Epoch 1/1 301/301 [==============================] - 6902s 23s/step - loss: 0.0413 - Dec_GT_Output_loss: -0.0065 - Dec_VAE_Output_loss: 0.0478 - Dec_GT_Output_dice_coefficient: 0.0065 - Dec_VAE_Output_dice_coefficient: 0.7997 - val_loss: 0.0229 - val_Dec_GT_Output_loss: 0.0000e+00 - val_Dec_VAE_Output_loss: 0.0229 - val_Dec_GT_Output_dice_coefficient: 0.0000e+00 - val_Dec_VAE_Output_dice_coefficient: 0.9005

Can some one help me..?

are you able to solve this problem?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/IAmSuyogJadhav/3d-mri-brain-tumor-segmentation-using-autoencoder-regularization/issues/33?email_source=notifications&email_token=AJYY3NSGNG2Z4E5TJWXQWYLRCZXRFA5CNFSM4KO6X44KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOELYLYPA#issuecomment-586202172, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJYY3NRTGUAHBN6GACIEDPDRCZXRFANCNFSM4KO6X44A .

-- With regards, Vijay Annepu, Regd No:161FA04281, phn : 8309524018. CSE Department, VIGNAN'S UNIVERSITY.

sneh-debug commented 4 years ago

Epoch 33/90 1/2 [==============>...............] - ETA: 15s - loss: 0.0630 - Dec_GT_Output_loss: 0.0032 - Dec_VAE_Output_loss: 0.0598 - Dec_GT_Output_dice_coefficient: 0.002/2 [==============================] - 31s 15s/step - loss: 0.0617 - Dec_GT_Output_loss: 0.0025 - Dec_VAE_Output_loss: 0.0592 - Dec_GT_Output_dice_coefficient: 0.0025 - Dec_VAE_Output_dice_coefficient: 0.7253 Dec_GT_Output_loss: 0.0025 & Dec_GT_Output_dice_coefficient: 0.0025 values are same at each iteration..