CODEWITHSUNDEEP
kerascomputer-visionefficientnet
cris tiano
cris tiano

Reputation: 1

why the output of EfficientNet B0 is two-dimension?

i know from paper:the output of efficientnet b0 is (*,7,7,1280),right?if so,the globalAveragePooling2D will get ndim = 4,instead of 2.

        model=Sequential()
        inputS=(height,width,depth)
        chanDim=-1
        model.add(EfficientNetB0(inputS, include_top=True, weights='imagenet'))
        model.add(GlobalAveragePooling2D())
        model.add(Dense(1024))
        model.add(Activation("swish"))
        model.add(BatchNormalization(axis=chanDim))
        model.add(Dropout(0.25))
        model.add(Dense(256))
        model.add(Activation("swish"))
        model.add(BatchNormalization(axis=chanDim))
        model.add(Dropout(0.25))
        model.add(Dense(32))
        model.add(Activation("tanh"))
        model.add(BatchNormalization(axis=chanDim))
        model.add(Dropout(0.25))
        model.add(Dense(classes))
        model.add(Activation("softmax"))
        return model

ValueError: Input 0 is incompatible with layer global_average_pooling2d_2: expected ndim=4, found ndim=2

Upvotes: 0

Views: 1586

Answers (1)

Dr. Snoopy
Dr. Snoopy

Reputation: 56367

That is because you set include_top to True, meaning that classification layers are included in the model, so the output shape of the whole model is (samples, classes), that's probably not what you want.

As you want feature maps, you should set include_top to False in the instantiation of EfficientNetB0.

Upvotes: 1

Related Questions