Dear author:
when I run "fine-tune.py",Error accurred. Please help me .I tried to replace "categorical_crossentropy" with "sparse_categorical_crossentropy",but nothing.

ValueError: You are passing a target array of shape (32, 1) while using as loss categorical_crossentropy. categorical_crossentropy expects targets to be binary matrices (1s and 0s) of shape (samples, classes). If your targets are integer classes, you can convert them to the expected format via:

from keras.utils.np_utils import to_categorical
y_binary = to_categorical(y_int)

Alternatively, you can use the loss function sparse_categorical_crossentropy instead, which does expect integer targets.

Hello, your transfer learning sample is the best I found in Internet, because it's easy to read (well coded) and reusable (using parameters). I used for some tests, it works perfectly.

I have a question now, do you think is it possible to make transfer learning from Inceptionv3, to add 2 new classes to the 1000 classes already classified from Imagenet?
For example I would like to create a model that is trained on 1000 images of Inceptionv3 + 2 classes of uncommon animals.

Thanks a lot.

Hi,

training worked very well - however when I run
python predict.py --model new.model --image images\20170530_200958897_iOS.jpg

i get this error :
File "predict.py", line 70, in <module>
ValueError: incompatible sizes: argument 'width' must be length 2 or scalar

any idea how to fix that ?

I substituted several imagenet models in the classify test , which all worked, except the densenet ones.

I also pulled down a keras-squeezenet from this link, and it worked after a small update to the _obtain_input_shape import and call. I posted the details on their issues.

https://github.com/wohlert/keras-squeezenet

I was wondering what is this imagenet vocabulary model readable terms.
I found some online, but unsure the order if it matches those. Thank you