ykamikawa / tf-keras-pspnet Goto Github PK

Hi!
First of all, thank you very much for sharing your code.
I would like to know if I could apply this network to a single channel MR images.

Thank you very much!

Hi,

How can I fine tune the model on a custom dataset? do you have the pre-trained model for download?

where is predict.py?

I used your model train my data but got a error:

ValueError: Error when checking target: expected activation_51 to have shape (None, 230400, 2) but got array with shape (998, 480, 480)

Can you help me solve it ?

Your mask resizing is changing the pixel values (due to default interpolation).
For example:
print(np.unique(original_mask))
[ 0 2 6 7 10]
print(np.unique(resized_mask))
[ 0 1 2 3 4 5 6 7 8 9 10]

You should use: "interpolation = cv2.INTER_NEAREST"
resized_mask = cv2.resize(original_mask, (dims[0], dims[1]), interpolation = cv2.INTER_NEAREST)
print(np.unique(original_mask))
[ 0 2 6 7 10]
print(np.unique(resized_mask))
[ 0 2 6 7 10]