在valid情況下，  輸出形狀計算方法為： new_height=new_width=[(W-F+1)/S]

在same情況下，輸出形狀計算方法為： new_height=new_width=[W/S]

其中W為輸入的尺寸，F為濾波器尺寸，S為步長，[ ]為向上取整函式。

由上可知，如果要保持卷積或池化之後的影象尺寸不變，則步長S必須為1。

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

# load資料
# import input_data
# mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)

# 定義網路超引數
learning_rate = 0.001
training_iters = 130000
batch_size = 64
display_step = 20

# 定義網路引數
n_input = 784 # 輸入的維度
n_classes = 10 # 標籤的維度
dropout = 0.8 # Dropout 的概率

# 佔位符輸入
with tf.variable_scope('Input'):
    x = tf.placeholder(tf.float32, [None, n_input])
    y = tf.placeholder(tf.float32, [None, n_classes])
    keep_prob = tf.placeholder(tf.float32)

# 卷積操作
def conv2d(name, l_input, w, b):
    return tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(l_input, w, strides=[1, 1, 1, 1], padding='SAME'),b), name=name)

# 最大下采樣操作
def max_pool(name, l_input, k1, k2):
    return tf.nn.max_pool(l_input, ksize=[1, k1, k1, 1], strides=[1, k2, k2, 1], padding='SAME', name=name)

# 歸一化操作
def norm(name, l_input, lsize=4):
    return tf.nn.lrn(l_input, lsize, bias=1.0, alpha=0.001 / 9.0, beta=0.75, name=name)

# 定義整個網路
def alex_net(_X, _weights, _biases, _dropout):
    # 向量轉為矩陣
    _X = tf.reshape(_X, shape=[-1, 28, 28, 1])

    # 卷積層
    with tf.variable_scope('Conv1'):
        conv1 = conv2d('conv1', _X, _weights['wc1'], _biases['bc1']) #28*28
        # 下采樣層
        pool1 = max_pool('pool1', conv1, k1=3, k2=2)#14*14
        # 歸一化層
        norm1 = norm('norm1', pool1, lsize=4)
        # Dropout
        norm1 = tf.nn.dropout(norm1, _dropout)
        tf.summary.histogram('conv', conv1)
        tf.summary.histogram('norm', norm1)


    # 卷積
    with tf.variable_scope('Conv2'):
        conv2 = conv2d('conv2', norm1, _weights['wc2'], _biases['bc2'])#14*14
        # 下采樣
        pool2 = max_pool('pool2', conv2, k1=3, k2=2)#7*7
        # 歸一化
        norm2 = norm('norm2', pool2, lsize=4)
        # Dropout
        norm2 = tf.nn.dropout(norm2, _dropout)
        tf.summary.histogram('conv', conv2)
        tf.summary.histogram('norm', norm2)

    # 卷積
    with tf.variable_scope('Conv3'):
        conv3 = conv2d('conv3', norm2, _weights['wc3'], _biases['bc3'])#7*7
        # 下采樣
        pool3 = max_pool('pool3', conv3, k1=3, k2=2) #4*4
        # 歸一化
        norm3 = norm('norm3', pool3, lsize=4)
        # Dropout
        norm3 = tf.nn.dropout(norm3, _dropout)
        tf.summary.histogram('conv', conv3)
        tf.summary.histogram('norm', norm3)

    # 全連線層，先把特徵圖轉為向量
    with tf.variable_scope('FC'):
        dense1 = tf.reshape(norm3, [-1, _weights['wd1'].get_shape().as_list()[0]])
        dense1 = tf.nn.relu(tf.matmul(dense1, _weights['wd1']) + _biases['bd1'], name='fc1')
        # 全連線層
        dense2 = tf.nn.relu(tf.matmul(dense1, _weights['wd2']) + _biases['bd2'], name='fc2') # Relu activation
        tf.summary.histogram('fc_out', dense2)

    # 網路輸出層
    with tf.variable_scope('Out'):
        out = tf.matmul(dense2, _weights['out']) + _biases['out']
        tf.summary.histogram('pred', out)
        return out

# 儲存所有的網路引數
weights = {
    'wc1': tf.Variable(tf.random_normal([3, 3, 1, 64])),
    'wc2': tf.Variable(tf.random_normal([3, 3, 64, 128])),
    'wc3': tf.Variable(tf.random_normal([3, 3, 128, 256])),
    'wd1': tf.Variable(tf.random_normal([4*4*256, 1024])),
    'wd2': tf.Variable(tf.random_normal([1024, 1024])),
    'out': tf.Variable(tf.random_normal([1024, 10]))
}

biases = {
    'bc1': tf.Variable(tf.random_normal([64])),
    'bc2': tf.Variable(tf.random_normal([128])),
    'bc3': tf.Variable(tf.random_normal([256])),
    'bd1': tf.Variable(tf.random_normal([1024])),
    'bd2': tf.Variable(tf.random_normal([1024])),
    'out': tf.Variable(tf.random_normal([n_classes]))
}


# 構建模型
pred = alex_net(x, weights, biases, keep_prob)

# 定義損失函式和學習步驟
with tf.variable_scope('Cost'):
    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred,labels= y))
    tf.summary.scalar('cost',cost)
with tf.variable_scope('Train'):
    optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)

# 測試網路
with tf.variable_scope('Accuracy'):
    correct_pred = tf.equal(tf.argmax(pred,1), tf.argmax(y,1))
    accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
    tf.summary.scalar('Acc',accuracy)

# 初始化所有的共享變數
init = tf.initialize_all_variables()

# 開啟一個訓練
with tf.Session() as sess:
    sess.run(init)

    writer = tf.summary.FileWriter('./log', sess.graph)  # write to file
    merge_op = tf.summary.merge_all()  # operation to merge all summary

    step = 1
    # Keep training until reach max iterations
    while step * batch_size < training_iters:
        batch_xs, batch_ys = mnist.train.next_batch(batch_size)
        # 獲取批資料
       # sess.run(optimizer, feed_dict={x: batch_xs, y: batch_ys, keep_prob: dropout})
        _, result = sess.run([optimizer, merge_op], {x: batch_xs, y: batch_ys, keep_prob: dropout})
        writer.add_summary(result, step) #record to file

        if step % display_step == 0:
            # 計算精度
            acc = sess.run(accuracy, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 1.})
            # 計算損失值
            loss = sess.run(cost, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 1.})
            print ("Iter " + str(step*batch_size) + ", Minibatch Loss= " + "{:.6f}".format(loss) + ", Training Accuracy= " + "{:.5f}".format(acc))
        step += 1

    print ("Optimization Finished!")
    # 計算測試精度
    print ("Testing Accuracy:", sess.run(accuracy, feed_dict={x: mnist.test.images[:256], y: mnist.test.labels[:256], keep_prob: 1.}))