squeezenet-keras

squeezenet-keras copied to clipboard

 due to stackoverflow Keras 1.2 doesn't support separate kernel/bias initializers, layers only take a single init parameter, https://stackoverflow.com/questions/59092788/typeerror-keyword-argument-not-understood-init-in-classifier-add I think we need to adapt the code for new keras...

mcagriaksoy

Running on Keras 1.0 docker (gw000/keras:1.2.2) running Tensorflow 1.0. I'm getting: ValueError: Negative dimension size caused by subtracting 3 from 2 for 'MaxPool' (op: 'MaxPool') with input shapes: [?,2,112,96].

scchess

If you are facing this error. Try to use this solution ```python tf.keras.layers.Concatenate(axis=-1, **kwargs) ```

ashishpatel26

no reduction in loss

1

The loss during training is not reducing. loss is constant at around 0.69 and and accuracy at around 0.5

amritbhaskar

How can I preprocess data to use your code?? Just by downloading 'Cats&Dog' and run your code occur error. Thanks.

DonghoonPark12

Hello, Thank you for sharing your code on Github. It would be of great help to me if I could get access to weights.h5 file generated after training. It would...

muthiyanbhushan