dingchenwei / googlenet

here is my implementation of the basic googLeNet describe in the paper "Going Deeper with Convolutions" in keras

googLeNet

here is my implementation of the basic googLeNet described in the paper Going Deeper with Convolutions in keras

I wrapped the inception module described in the paper and put it in the file "inceptionModel.py"

the structure of the net is described as following:

googlenet's Issues

wrong way in adding flatten layer

In googLeNet.py line 49:
linear = Dense(units=1000, activation="softmax", `kernel_regularizer=l2(0.01))(keras.layers.core.Flatten(drop1))
while it should be:
linear = Dense(units=1000, activation="softmax", kernel_regularizer=l2(0.01))(keras.layers.core.Flatten()(drop1))

missing flatten layer

Hi, I believe there is a missing Flatten() layer between 'drop1' and 'linear'