Different output shape at last layer and the predict

0
votes

I have build a Keras sequencial model using a pretrained model with addition of some layers. The code is as below. Then tried to predict and expected prediction shape was number of (samples,16) but got the prrediction results as (samples,8). The code for model building and the shape printing are as below.

`

layer_output = base_model.get_layer(output_layer).output
x=layer_output
x = Dense(1024, activation='sigmoid',name='1024_out')(x)
x = Dense(512, activation='sigmoid', name='512_out')(x)
x = Dense(16, activation='sigmoid',name="final")(x)
model = Model(base_model.input, outputs=x)
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='mean_squared_error', optimizer=sgd)
model.load_weights(weight_path)

The prediction shape andd the shape of last layer output print("Pred ",model.predict(images[:2]).shape,"Last Layer:",model.layers[-1].output_shape)` The output is Pred (2, 8) Last Layer: (None, 16)

1
pythontensorflowkeras

1 Answers

1
votes

Got the answer to the above question/issue. The last layer was not sigmoid instead of softmax for the multi-class classification. Changed 

x = Dense(16, activation='sigmoid',name="final")(x)

to

x = Dense(16, activation='softmax',name="final")(x)