tfserve_vision7.py

INPUT_SHAPE_RN = (32, 32, 3)

def create_cnn_architecture_model2(input_shape):
    inc_net = keras.applications.resnet50.ResNet50(include_top=False, weights='imagenet', 
                                                          input_shape=input_shape)
    inc_net.trainable = True
    # Fine-tune the layers
    for layer in inc_net.layers:
        layer.trainable = True
    
    base_inc = inc_net
    base_out = base_inc.output
    pool_out = keras.layers.Flatten()(base_out)
    
    hidden1 = keras.layers.Dense(512, activation='relu')(pool_out)
    drop1 = keras.layers.Dropout(rate=0.3)(hidden1)
    hidden2 = keras.layers.Dense(512, activation='relu')(drop1)
    drop2 = keras.layers.Dropout(rate=0.3)(hidden2)

    out = keras.layers.Dense(10, activation='softmax')(drop2)

    model = keras.Model(inputs=base_inc.input, outputs=out)
    model.compile(optimizer=keras.optimizers.RMSprop(lr=1e-4),
                    loss='sparse_categorical_crossentropy',
                    metrics=['accuracy'])
    return model
    
model2 = create_cnn_architecture_model2(input_shape=INPUT_SHAPE_RN)
model2.summary()

# Output
Model: "model_3"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_4 (InputLayer)            [(None, 32, 32, 3)]  0                                            
__________________________________________________________________________________________________
conv1_pad (ZeroPadding2D)       (None, 38, 38, 3)    0           input_4[0][0]                    
__________________________________________________________________________________________________
conv1 (Conv2D)                  (None, 16, 16, 64)   9472        conv1_pad[0][0]                  
__________________________________________________________________________________________________
...
...
dropout_5 (Dropout)             (None, 512)          0           dense_8[0][0]                    
__________________________________________________________________________________________________
dense_9 (Dense)                 (None, 10)           5130        dropout_5[0][0]                  
==================================================================================================
Total params: 24,904,586
Trainable params: 24,851,466
Non-trainable params: 53,120
__________________________________________________________________________________________________