如何在 TensorFlow 中查看层的输出大小？答案

【问题标题】：How to see the output size of a layer in TensorFlow?如何在 TensorFlow 中查看层的输出大小？
【发布时间】：2018-02-12 19:42:06
【问题描述】：

我是 TensorFlow 的新手，目前正在使用该库编写我的第一个 CNN。以前我使用过 keras 并使用 model.summary() 函数检查图层的输出尺寸。如何检查 TensorFlow 中层的输出尺寸？这是我的模型：

generator(input, random_dim, is_train, reuse=False):
    c4, c8, c16, c32, c64 = 512, 256, 128, 64, 32 # no of nodes in each conv2d layer
    s4 = 4
    output_dim = CHANNEL  # B/W image for RGB channel = 3
    with tf.variable_scope('gen') as scope: 
        if reuse:
            scope.reuse_variables()
        w1 = tf.get_variable('w1', shape=[random_dim, s4 * s4 * c4], dtype=tf.float32,
                             initializer=tf.truncated_normal_initializer(stddev=0.02))
        b1 = tf.get_variable('b1', shape=[c4 * s4 * s4], dtype=tf.float32,
                             initializer=tf.constant_initializer(0.0))
        flat_conv1 = tf.add(tf.matmul(input, w1), b1, name='flat_conv1')

        conv1 = tf.reshape(flat_conv1, shape=[-1, s4, s4, c4], name='conv1')
        bn1 = tf.contrib.layers.batch_norm(conv1, is_training=is_train, epsilon=1e-5, decay = 0.9,  updates_collections=None, scope='bn1')
        act1 = tf.nn.relu(bn1, name='act1')
        # 8*8*256

        conv2 = tf.layers.conv2d_transpose(act1, c8, kernel_size=[3, 3], strides=[2, 2], padding="SAME",
                                           kernel_initializer=tf.truncated_normal_initializer(stddev=0.02),
                                           name='conv2')
        bn2 = tf.contrib.layers.batch_norm(conv2, is_training=is_train, epsilon=1e-5, decay = 0.9,  updates_collections=None, scope='bn2')
        act2 = tf.nn.relu(bn2, name='act2')
        # 16*16*128
        conv3 = tf.layers.conv2d_transpose(act2, c16, kernel_size=[3, 3], strides=[2, 2], padding="SAME",
                                           kernel_initializer=tf.truncated_normal_initializer(stddev=0.02),
                                           name='conv3')
        bn3 = tf.contrib.layers.batch_norm(conv3, is_training=is_train, epsilon=1e-5, decay = 0.9,  updates_collections=None, scope='bn3')
        act3 = tf.nn.relu(bn3, name='act3')
        # 32*32*64
        conv4 = tf.layers.conv2d_transpose(act3, c32, kernel_size=[3, 3], strides=[2, 2], padding="SAME",
                                           kernel_initializer=tf.truncated_normal_initializer(stddev=0.02),
                                           name='conv4')
        bn4 = tf.contrib.layers.batch_norm(conv4, is_training=is_train, epsilon=1e-5, decay = 0.9,  updates_collections=None, scope='bn4')
        act4 = tf.nn.relu(bn4, name='act4')
        # 64*64*32
        conv5 = tf.layers.conv2d_transpose(act4, c64, kernel_size=[3, 3], strides=[2, 2], padding="SAME",
                                           kernel_initializer=tf.truncated_normal_initializer(stddev=0.02),
                                           name='conv5')
        bn5 = tf.contrib.layers.batch_norm(conv5, is_training=is_train, epsilon=1e-5, decay = 0.9,  updates_collections=None, scope='bn5')
        act5 = tf.nn.relu(bn5, name='act5')

        #128*128*3
        conv6 = tf.layers.conv2d_transpose(act5, output_dim, kernel_size=[3, 3], strides=[2, 2], padding="SAME",
                                           kernel_initializer=tf.truncated_normal_initializer(stddev=0.02),
                                           name='conv6')
        # bn6 = tf.contrib.layers.batch_norm(conv6, is_training=is_train, epsilon=1e-5, decay = 0.9,  updates_collections=None, scope='bn6')

        #BATCH NORM IN EVERY LAYER EXCEPT LAST !
        act6 = tf.nn.tanh(conv6, name='act6')
        return act6

使用 python 3

【问题讨论】：

试试print(conv6.get_shape())，其中conv6是你感兴趣的张量。
@MaosiChen 我认为您应该发表您的评论作为答案。

标签： python tensorflow deep-learning conv-neural-network

【解决方案1】：

在编译时（未知维度将具有None 或? 值）：

a = ...  # Your tensor
print(a.shape.dims)

在运行时（未知维度将从输入数据计算）：

sess = tf.Session()
a = ... # Your tensor
feed_dict = {...}  # Values required to compute a
shape_op = tf.shape(a)
shape_res = sess.run(shape_op, feed_dict=feed_dict)
print(shape_res)

干杯

【讨论】：