第二个卷积层的输入维数？

Question

In the following code I have introduced one convolution and one maximum pooling layer. Output of pooling layer has a shape of (4,6,6,1). Now I want to define second convolutional layer. What will be the input for the second convolutional layer? Can I call same conv2d function? But here the input channel is different.

 batch_size = 4
 image_height = 12
 image_width =12
 input_channel = 2
 output_channel =1
 input =  tf.Variable(tf.random_normal([batch_size,image_height,image_width,input_channel]))
 filter = tf.Variable(tf.random_normal([2,2,input_channel,output_channel]))
 def conv2d(inputs,filters):
     return tf.nn.conv2d(inputs,filters,strides=[1,1,1,1],padding='SAME')
 def max_pool(conv_out):
     return tf.nn.max_pool(conv_out,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')
 conv_out1 = conv2d(input,filter)
 pooling_out1= max_pool(conv_out1)
 sess =tf.InteractiveSession()
 sess.run(tf.initialize_all_variables())
 print conv_out1.get_shape()
 print pooling_out1.get_shape()

Answer 1

您可以调用相同的conv2d函数，但希望使用不同的过滤器。