tensorflow练习12：利用图片预测年龄与性别

深度学习在图像分类领域已经取得长足地进展，以下以一个有趣的例子来学习图像分类算法。
训练数据：人脸数据集（链接: https://pan.baidu.com/s/1gf4FQD1 密码: ddkx）
环境:tensorflow,python3.5
1)加载数据集并对数据进行处理

age_table = ['(0, 2)', '(4, 6)', '(8, 12)', '(15, 20)', '(25, 32)', '(38, 43)', '(48, 53)', '(60, 100)']sex_table = ['f', 'm']#性别# AGE==True 训练年龄模型，False,训练行性别模型AGE = Trueif AGE == True:    labels_size = len(age_table)  # 年龄else:    labels_size = len(sex_table)  # 性别face_set_fold = 'AdienceBenchmarkOfUnfilteredFacesForGenderAndAgeClassification'#加载图片目录文件fold_0_data = os.path.join(face_set_fold, 'fold_0_data.txt')fold_1_data = os.path.join(face_set_fold, 'fold_1_data.txt')fold_2_data = os.path.join(face_set_fold, 'fold_2_data.txt')fold_3_data = os.path.join(face_set_fold, 'fold_3_data.txt')fold_4_data = os.path.join(face_set_fold, 'fold_4_data.txt')face_image_set = os.path.join(face_set_fold, 'aligned')#图片数据库def parse_data(fold_x_data):    data_set = []    with open(fold_x_data, 'r') as f:        line_one = True        for line in f:            tmp = []            if line_one == True:                line_one = False                continue#舍弃第一行标题            tmp.append(line.split('\t')[0])#user_id            tmp.append(line.split('\t')[1])#xx.jpg            tmp.append(line.split('\t')[3])#年龄            tmp.append(line.split('\t')[4])#性别            file_path = os.path.join(face_image_set, tmp[0])#目录级            if os.path.exists(file_path):                #返回所有匹配的文件路径列表。它只有一个参数pathname，定义了文件路径匹配规则                filenames = glob.glob(file_path + "/*.jpg")#返回所有图片的列表                for filename in filenames:                    if tmp[1] in filename:                        break                if AGE == True:                    if tmp[2] in age_table:                        data_set.append([filename, age_table.index(tmp[2])])#文件名，年龄表                else:                    if tmp[3] in sex_table:                        data_set.append([filename, sex_table.index(tmp[3])])    return data_set#返回数据集data_set_0 = parse_data(fold_0_data)data_set_1 = parse_data(fold_1_data)data_set_2 = parse_data(fold_2_data)data_set_3 = parse_data(fold_3_data)data_set_4 = parse_data(fold_4_data)data_set = data_set_0 + data_set_1 + data_set_2 + data_set_3 + data_set_4#数据集串联shuffle(data_set)#随机排序

根据文件中的数据，处理得到需要用到的数据；（注：解压aligned文件）

2）定义数据结构

# 缩放图像大小IMAGE_WIDTH = 227IMAGE_HEIGHT = 227# 读取缩放图像jpg_data = tf.placeholder(dtype=tf.string)docode_jpg = tf.image.decode_jpeg(jpg_data, channels=3)#Returns:A `Tensor` of type `uint8`. 3-D with shape `[height, width, channels]`.resize = tf.image.resize_images(docode_jpg, [IMAGE_HEIGHT, IMAGE_WIDTH])resize = tf.cast(resize, tf.uint8) / 255def resize_image(file_name):    with tf.gfile.FastGFile(file_name, 'r') as f:        image_data = f.read()    with tf.Session() as sess:        image = sess.run(resize, feed_dict={jpg_data: image_data})    return imagepointer = 0#指针指示def get_next_batch(data_set, batch_size=128):#获取一个batch数据    global pointer    batch_x = []    batch_y = []    for i in range(batch_size):        batch_x.append(resize_image(data_set[pointer][0]))        batch_y.append(data_set[pointer][1])        pointer += 1    return batch_x, batch_ybatch_size = 128num_batch = len(data_set) // batch_sizeX = tf.placeholder(dtype=tf.float32, shape=[batch_size, IMAGE_HEIGHT, IMAGE_WIDTH, 3])Y = tf.placeholder(dtype=tf.int32, shape=[batch_size])

3）定义网络结构，3层卷积、池化，2层卷积；

def conv_net(nlabels, images, pkeep=1.0):    """定义卷积网络的结构"""    weights_regularizer = tf.contrib.layers.l2_regularizer(0.0005)    # 默认变量域为con_net    with tf.variable_scope("conv_net", "conv_net", [images]) as scope:        # 给卷积层和全连接层设置默认参数        with tf.contrib.slim.arg_scope([convolution2d, fully_connected], weights_regularizer=weights_regularizer,                                       biases_initializer=tf.constant_initializer(1.),                                       weights_initializer=tf.random_normal_initializer(stddev=0.005), trainable=True):            # 给卷积层权重设置默认参数            with tf.contrib.slim.arg_scope([convolution2d],                                           weights_initializer=tf.random_normal_initializer(stddev=0.01)):                #卷积层1：滤波器（7*7）,步长4，输出滤波器数：96                conv1 = convolution2d(images, 96, [7, 7], [4, 4], padding='VALID',                                      biases_initializer=tf.constant_initializer(0.), scope='conv1')                pool1 = max_pool2d(conv1, 3, 2, padding='VALID', scope='pool1')                norm1 = tf.nn.local_response_normalization(pool1, 5, alpha=0.0001, beta=0.75, name='norm1')                #卷积层2：                conv2 = convolution2d(norm1, 256, [5, 5], [1, 1], padding='SAME', scope='conv2')                pool2 = max_pool2d(conv2, 3, 2, padding='VALID', scope='pool2')                norm2 = tf.nn.local_response_normalization(pool2, 5, alpha=0.0001, beta=0.75, name='norm2')                #卷积层3：                conv3 = convolution2d(norm2, 384, [3, 3], [1, 1], biases_initializer=tf.constant_initializer(0.),                                      padding='SAME', scope='conv3')                pool3 = max_pool2d(conv3, 3, 2, padding='VALID', scope='pool3')                flat = tf.reshape(pool3, [-1, 384 * 6 * 6], name='reshape')                #两层全连接                full1 = fully_connected(flat, 512, scope='full1')                drop1 = tf.nn.dropout(full1, pkeep, name='drop1')                full2 = fully_connected(drop1, 512, scope='full2')                drop2 = tf.nn.dropout(full2, pkeep, name='drop2')    #输出域    with tf.variable_scope('output') as scope:        weights = tf.Variable(tf.random_normal([512, nlabels], mean=0.0, stddev=0.01), name='weights')        biases = tf.Variable(tf.constant(0.0, shape=[nlabels], dtype=tf.float32), name='biases')        output = tf.add(tf.matmul(drop2, weights), biases, name=scope.name)    return output

4)训练

def training():#训练    logits = conv_net(labels_size, X)#经过卷积、池化、全连接后的结果    def optimizer(eta, loss_fn):        global_step = tf.Variable(0, trainable=False)        optz = lambda lr: tf.train.MomentumOptimizer(lr, 0.9)        lr_decay_fn = lambda lr, global_step: tf.train.exponential_decay(lr, global_step, 100, 0.97, staircase=True)        return tf.contrib.layers.optimize_loss(loss_fn, global_step, eta, optz, clip_gradients=4.,                                               learning_rate_decay_fn=lr_decay_fn)    def loss(logits, labels):        cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits, labels)        cross_entropy_mean = tf.reduce_mean(cross_entropy)        regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)        total_loss = cross_entropy_mean + 0.01 * sum(regularization_losses)        loss_averages = tf.train.ExponentialMovingAverage(0.9)        loss_averages_op = loss_averages.apply([cross_entropy_mean] + [total_loss])        with tf.control_dependencies([loss_averages_op]):            total_loss = tf.identity(total_loss)        return total_loss    # loss    total_loss = loss(logits, Y)    # optimizer    train_op = optimizer(0.001, total_loss)    saver = tf.train.Saver(tf.global_variables())    with tf.Session() as sess:        sess.run(tf.global_variables_initializer())        print("开始训练:")        print(time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()))        global pointer        epoch = 0        while epoch < 51:#50个epoch            pointer = 0            for batch in range(num_batch):                batch_x, batch_y = get_next_batch(data_set, batch_size)#获取数据                _, loss_value = sess.run([train_op, total_loss], feed_dict={X: batch_x, Y: batch_y})                print(time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()))                print("%d th epoch %d th batch loss:%f "%(epoch+1, batch+1, loss_value))            i = 1            if epoch % 5 == 0:                print("第%d次保存模型" % i)                print(time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()))                saver.save(sess, './age.module' if AGE == True else './sex.module', global_step=epoch)                epoch += 1                i += 1training()