Closed EderSantana closed 7 years ago
recent versions of keras now uses Tensorflow b01c dim ordering as default, this sets the ordering back to theano's, which is what the codebase is adapted for.
This can also be solved by modifying the keras.json file in .keras mine had as default
"image_dim_ordering": "tf",
I just changed it back to th
yup
Hi Eder, sorry for the long delay in this and thanks a lot for your help. I've made it now so that it can handle both image_dim_orderings and backends.
Fix #3 for recent versions of keras