使用tensorflow訓練自己的資料集—定義神經網路

上一篇使用tensorflow訓練自己的資料集（一）中製作已經介紹了製作自己的資料集、接下來就是定義向前傳播過程了也就是定義神經網路。本次使用了兩層卷積兩層最大池化兩層全連線神經網路最後加softmax層的經典卷積神經網路結構。

import tensorflow as tf
# 配置引數
# 圖片size
IMAGE_SIZE = 128
NUM_CHANNELS = 3
NUM_LABELS = 2
# 第一層卷積層的尺寸和深度
CONV1_DEEP = 64
CONV1_SIZE = 5
# 第二層卷積層的尺寸和深度
CONV2_DEEP =
 
 128
CONV2_SIZE = 5
# 全連線層的節點個數
FC_SIZE = 512
def get_Weight(shape,regularizer_rate = None):     # 定義weight如需正則化需傳入zhengzehualv預設值為None
    Weight = tf.Variable(tf.truncated_normal(shape=shape,stddev=0.1),dtype=tf.float32)    # tensorflow API推薦隨機初始化

    if regularizer_rate != None:
        regularizer =
 
 tf.contrib.layers.l2_regularizer(regularizer_rate)
        tf.add_to_collection('losses',regularizer(Weight))

    return Weight

def get_biase(shape):       # 定義biase
    biase = tf.Variable(tf.constant(value=0.1,shape=shape),dtype=tf.float32)    # tensorflow API推薦初始化0.1
    return biase

def create_conv2d
 
(x,w):     # 定義卷積層
    conv2d = tf.nn.conv2d(x,w,strides=[1,1,1,1],padding='SAME')     # 步幅為1、SAME填充
    return conv2d

def max_pooling(x):         # 定義最大值池化
    pool = tf.nn.max_pool(x,ksize=[1,2,2,1],strides=[1,2,2,1],padding='SAME')   # ksize為2、步幅為2、SAME填充
    return pool

def create_fc(x,w,b):       # 定義全連線層
    fc = tf.matmul(x,w) + b
    return fc

# 定義前向傳播的過程
# 這裡添加了一個新的引數train，用於區分訓練過程和測試過程。
# 在這個程式中將用到dropout方法，dropout可以進一步提升模型可靠性並防止過擬合
# dropout過程只在訓練時使用
def inference(input_tensor, train, regularizer_rate):
    with tf.variable_scope('layer1-conv1'):
        conv1_Weights = get_Weight([CONV1_SIZE,CONV1_SIZE,NUM_CHANNELS,CONV1_DEEP])     # 5*5*64
        conv1_baises = get_biase([CONV1_DEEP])
        conv1 = tf.nn.bias_add(create_conv2d(input_tensor,conv1_Weights),conv1_baises)
        conv1 = tf.nn.relu(conv1)       # 使用ReLu啟用函式

    with tf.name_scope('layer2-pool1'):        # 64*64*64
        pool1 = max_pooling(conv1)

    with tf.variable_scope('layer3-conv2'):
        conv2_Weights = get_Weight([CONV2_SIZE,CONV2_SIZE,CONV1_DEEP,CONV2_DEEP])       # 5*5*128
        conv2_biases = get_biase([CONV2_DEEP])
        conv2 = tf.nn.bias_add(create_conv2d(pool1,conv2_Weights),conv2_biases)
        conv2 = tf.nn.relu(conv2)

    with tf.name_scope('layer4-pool2'):         # 32*32*128
        pool2 = max_pooling(conv2)

        pool_shape = pool2.get_shape().as_list()
        # pool_shape為[batch_size,32,32,128]
        # 計算將矩陣拉直成向量之後的長度，這個長度就是矩陣長度及深度的乘積。
        nodes = pool_shape[1] * pool_shape[2] * pool_shape[3]
        # 通過tf.reshape函式將第四層的輸出變成一個batch的向量
        reshaped = tf.reshape(pool2, [pool_shape[0], nodes])

    # 宣告第五層全連線層的變數並實現前向傳播過程
    with tf.variable_scope('layer5-fc1'):
        fc1_Weights = get_Weight([nodes,FC_SIZE],regularizer_rate)
        fc1_biases = get_biase([FC_SIZE])
        fc1 = tf.nn.relu(create_fc(reshaped,fc1_Weights,fc1_biases))
    # 訓練過程新增dropout防止過擬合
        if train:
            fc1 = tf.nn.dropout(fc1, 0.5)

    # 宣告第六層全連線層的變數並實現前向傳播過程
    with tf.variable_scope('layer6-fc2'):
        fc2_Weights = get_Weight([FC_SIZE,NUM_LABELS],regularizer_rate)
        fc2_biases = get_biase([NUM_LABELS])
        logit = create_fc(fc1,fc2_Weights,fc2_biases)
        # fc2 = tf.nn.relu(fc2)     需要softmax層時使用啟用函式

    # with tf.variable_scope('layer7-softmax'):
    # 可在全連線層後新增softmax層loss函式需要改變
    #     sm_Weight = get_Weight([FC_SIZE,NUM_LABELS])
    #     sm_biases = get_biase([NUM_LABELS])
    #     sm = tf.matmul(fc2,sm_Weight) + sm_biases
    #     logit = tf.nn.softmax(sm)
        return logit

下一篇將介紹反向傳播過程 如有錯誤望多多指教~~