In the second convolution layer, we start with the first layer's output as input and build a new layer with the following parameters:
First, we define a placeholder for real y and the class of real y (the label of the class):
y_true = tf.placeholder(tf.float32, shape=[None, num_classes], name='y_true') y_true_cls = tf.argmax(y_true, dimension=1)
The shape of these two variables is the following:
layer_conv2, weights_conv2 = new_conv_layer(input=layer_conv1, num_input_channels=num_filters1,filter_size=filter_size2,num_filters=num_filters2,use_pooling=True) ...