Convert to Tensorflow about 3dunetcnn HOT 7 CLOSED

keras is a wrapper for tensorflow, turn on tensorflow as backend for keras and start from there?

from 3dunetcnn.

The most noteworthy change is to switch the axis of concatenation to the last channel instead of the first. This is because Theano is channels_first but Tensorflow is channels_last. E.g. change 'concatenate([up5, conv3], axis=1)' to 'concatenate([up5, conv3], axis=4)'. Otherwise, things can generally stay the same.

from 3dunetcnn.

Could you please elaborate more on the changes needed to use TensorFlow as backend?
I have applied the fix above, setting axis=4, but am getting this message when trying to run:

ValueError: Error when checking input: expected input_1 to have shape (None, 64, 64, 32, 1) but got array with shape (1, 1, 64, 64, 32)

from 3dunetcnn.

I think it's from "data.py", function "add_data_to_storage". Any help would be appreciated.

from 3dunetcnn.

To use Tensorflow as a backend, you don't need to change any of the code. Simply follow the instructions listed in the Keras backend documentation.

from 3dunetcnn.

Tested with

KERAS_BACKEND=tensorflow python train.py

Works well

from 3dunetcnn.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. If you have questions, feel free to join the Slack group or email me at [email protected].

from 3dunetcnn.