基于传统神经网络，使用Keras训练自己的数据集

以图像多分类为例

1. 准备数据集：

以图像分类为例，准备了五种花的图片数据(每种各600张图片), 依次存放在'./flower_photos/daisy'、 './flower_photos/dandelion'、'./flower_photos/roses'、'./flower_photos/sunflowers'、'./flower_photos/tulips' 文件夹中。(其中，数据集及模型以上传,可自行下载）


2. 训练代码：

from skimage import io, transform
import glob
import os
import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np
import time

# 数据集地址
path = 'C:/Users/zhf/Desktop/flower/flower_photos/'
# 模型保存地址
model_path = 'model/model.ckpt'

# 将所有的图片resize成100*100
w = 100
h = 100
c = 3

# 读取图片
def read_img(path):
    cate = [path+x for x in os.listdir(path) if os.path.isdir(path+x)]
    imgs = []
    labels = []
    for idx, folder in enumerate(cate):   # idx代表的是图像的标签，随着依次遍历每个文件夹，在本程序中的数值为：0、1、2、3、4
        for im in glob.glob(folder+'/*.jpg'):  # glob.glob获取目录下的所有图片
            print('reading the images:%s' % (im))
            img=io.imread(im)  # 图像读取
            img=transform.resize(img,(w,h))  # 图像裁剪
            imgs.append(img)
            labels.append(idx)  # 标签满足
    return np.asarray(imgs, np.float32), np.asarray(labels, np.int32)
data, label = read_img(path)

# 打乱顺序
num_example = data.shape[0]   # 统计总共有3670张图片
arr = np.arange(num_example)  # [0 1 2...3669]
np.random.shuffle(arr)
data = data[arr]
label = label[arr]

# 将所有数据分为训练集和验证集
ratio = 0.8
s = np.int(num_example * ratio)
x_train = data[:s]
y_train = label[:s]
x_val = data[s:]
y_val = label[s:]

# -----------------构建网络----------------------
# tf.placeholder占位符,暂时储存变量.  在会话中，运行模型的时候通过feed_dict()函数向占位符送入数据
x = tf.placeholder(tf.float32, shape=[None, w, h, c], name='x')
y_ = tf.placeholder(tf.int32, shape=[None, ], name='y_')

def inference(input_tensor, train, regularizer):
    with tf.variable_scope('layer1-conv1'):   # tf.variable_scope实现变量命名空间，变量共享     # tf.truncated_normal_initializer按照正态分布
        conv1_weights = tf.get_variable("weight", [5, 5, 3, 32], initializer=tf.truncated_normal_initializer(stddev=0.1))  # 卷积核大小为5*5*3，暂时理解成在第一层进行了32次卷积操作，第一层有32个神经元
        conv1_biases = tf.get_variable("bias", [32], initializer=tf.constant_initializer(0.0))
        conv1 = tf.nn.conv2d(input_tensor, conv1_weights, strides=[1, 1, 1, 1], padding='SAME')  # 使用SAME时，发生不对应时，边界补0，不丢弃像素点
        relu1 = tf.nn.relu(tf.nn.bias_add(conv1, conv1_biases))  # 激活的过程

    with tf.name_scope("layer2-pool1"):
        pool1 = tf.nn.max_pool(relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="VALID")  # 发生不对应时，边界丢弃

    with tf.variable_scope("layer3-conv2"):
        conv2_weights = tf.get_variable("weight", [5, 5, 32, 64],initializer=tf.truncated_normal_initializer(stddev=0.1))
        conv2_biases = tf.get_variable("bias", [64], initializer=tf.constant_initializer(0.0))
        conv2 = tf.nn.conv2d(pool1, conv2_weights, strides=[1, 1, 1, 1], padding='SAME')
        relu2 = tf.nn.relu(tf.nn.bias_add(conv2, conv2_biases))

    with tf.name_scope("layer4-pool2"):
        pool2 = tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')

    with tf.variable_scope("layer5-conv3"):
        conv3_weights = tf.get_variable("weight",[3,3,64,128],initializer=tf.truncated_normal_initializer(stddev=0.1))
        conv3_biases = tf.get_variable("bias", [128], initializer=tf.constant_initializer(0.0))
        conv3 = tf.nn.conv2d(pool2, conv3_weights, strides=[1, 1, 1, 1], padding='SAME')
        relu3 = tf.nn.relu(tf.nn.bias_add(conv3, conv3_biases))

    with tf.name_scope("layer6-pool3"):
        pool3 = tf.nn.max_pool(relu3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')

    with tf.variable_scope("layer7-conv4"):
        conv4_weights = tf.get_variable("weight", [3, 3, 128, 128],initializer=tf.truncated_normal_initializer(stddev=0.1))
        conv4_biases = tf.get_variable("bias", [128], initializer=tf.constant_initializer(0.0))
        conv4 = tf.nn.conv2d(pool3, conv4_weights, strides=[1, 1, 1, 1], padding='SAME')
        relu4 = tf.nn.relu(tf.nn.bias_add(conv4, conv4_biases))

    with tf.name_scope("layer8-pool4"):
        pool4 = tf.nn.max_pool(relu4, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
        nodes = 6*6*128
        reshaped = tf.reshape(pool4,[-1,nodes])

    with tf.variable_scope('layer9-fc1'):
        fc1_weights = tf.get_variable("weight", [nodes, 1024],
                                      initializer=tf.truncated_normal_initializer(stddev=0.1))
        if regularizer != None: tf.add_to_collection('losses', regularizer(fc1_weights)) # tf.add_to_collection是把多个变量放入一个自己用引号命名的集合里，也就是把多个变量统一放在一个列表中。
        fc1_biases = tf.get_variable("bias", [1024], initializer=tf.constant_initializer(0.1))

        fc1 = tf.nn.relu(tf.matmul(reshaped, fc1_weights) + fc1_biases)
        if train: fc1 = tf.nn.dropout(fc1, 0.5)

    with tf.variable_scope('layer10-fc2'):
        fc2_weights = tf.get_variable("weight", [1024, 512],
                                      initializer=tf.truncated_normal_initializer(stddev=0.1))
        if regularizer != None: tf.add_to_collection('losses', regularizer(fc2_weights))
        fc2_biases = tf.get_variable("bias", [512], initializer=tf.constant_initializer(0.1))

        fc2 = tf.nn.relu(tf.matmul(fc1, fc2_weights) + fc2_biases)
        if train: fc2 = tf.nn.dropout(fc2, 0.5)

    with tf.variable_scope('layer11-fc3'):
        fc3_weights = tf.get_variable("weight", [512, 5],
                                      initializer=tf.truncated_normal_initializer(stddev=0.1))
        if regularizer != None: tf.add_to_collection('losses', regularizer(fc3_weights))
        fc3_biases = tf.get_variable("bias", [5], initializer=tf.constant_initializer(0.1))
        logit = tf.matmul(fc2, fc3_weights) + fc3_biases

    return logit

#---------------------------网络结束---------------------------
regularizer = tf.contrib.layers.l2_regularizer(0.0001)  # 正则化就是给网络加上一些规则，提高模型的适用性
logits = inference(x, False, regularizer)

#(小处理)将logits乘以1赋值给logits_eval，定义name，方便在后续调用模型时通过tensor名字调用输出tensor
b = tf.constant(value=1, dtype=tf.float32)
logits_eval = tf.multiply(logits, b, name='logits_eval')

loss = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, labels=y_)  # 损失函数的交叉熵
train_op = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)
correct_prediction = tf.equal(tf.cast(tf.argmax(logits,1),tf.int32), y_)
acc = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

# 定义一个函数，按批次取数据
def minibatches(inputs=None, targets=None, batch_size=None, shuffle=False):
    assert len(inputs) == len(targets)
    if shuffle:
        indices = np.arange(len(inputs))
        np.random.shuffle(indices)
    for start_idx in range(0, len(inputs) - batch_size + 1, batch_size):
        if shuffle:
            excerpt = indices[start_idx:start_idx + batch_size]
        else:
            excerpt = slice(start_idx, start_idx + batch_size)
        yield inputs[excerpt], targets[excerpt]

# 训练和测试数据，可将n_epoch设置更大一些

n_epoch = 10
batch_size = 64
saver = tf.train.Saver()
sess = tf.Session()
tf.summary.FileWriter("logs/", sess.graph)

sess.run(tf.global_variables_initializer())

loss_train = []
acc_train = []
loss_validation = []
acc_validation = []

for epoch in range(n_epoch):
    start_time = time.time()

    # training
    train_loss, train_acc, n_batch = 0.0, 0.0, 0.0
    for x_train_a, y_train_a in minibatches(x_train, y_train, batch_size, shuffle=True):
        _, err, ac = sess.run([train_op, loss, acc], feed_dict={x: x_train_a, y_: y_train_a}) # feed_dict中初始神经元的个数
        train_loss += err; train_acc += ac; n_batch += 1


    loss_train.append(np.sum(train_loss)/ n_batch)  # 正确理解 batch_size，n_batch以及n_epoch
    acc_train.append(np.sum(train_acc)/ n_batch)

    print("train loss: %f" % float(np.sum(train_loss)/ n_batch))
    print("train acc: %f" % float(np.sum(train_acc)/ n_batch))

    # validation
    val_loss, val_acc, n_batch = 0.0, 0.0, 0.0
    for x_val_a, y_val_a in minibatches(x_val, y_val, batch_size, shuffle=False):
        err, ac = sess.run([loss,acc], feed_dict={x: x_val_a, y_: y_val_a})
        val_loss += err; val_acc += ac; n_batch += 1

    loss_validation.append(np.sum(val_loss)/ n_batch)
    acc_validation.append(np.sum(val_acc)/ n_batch)

    print("validation loss: %f" % float(np.sum(val_loss)/ n_batch))
    print("validation acc: %f" % float(np.sum(val_acc)/ n_batch))
saver.save(sess,model_path)
sess.close()


# 绘制结果曲线
N = np.arange(0, n_epoch)
plt.style.use("ggplot")
plt.figure()
plt.plot(N, acc_train, label="train_acc",color="red")
plt.plot(N, acc_validation, label="val_acc", color="green")
plt.title("Training and Validation Accuracy (Simple NN)")
plt.xlabel("Epoch 10")
plt.ylabel("Accuracy")
plt.legend()
plt.savefig('simple_nn_plot_acc.png')

plt.figure()
plt.plot(N, loss_train, label="train_loss", color="red")
plt.plot(N, loss_validation, label="val_loss", color="green")
plt.title("Training and Validation Loss (Simple NN)")
plt.xlabel("Epoch #")
plt.ylabel("Loss")
plt.legend()
plt.savefig('simple_nn_plot_loss.png')

运行输出：

---

3.测试代码

from skimage import io,transform
import tensorflow as tf
import numpy as np

path2 = "flower_photos/daisy/5547758_eea9edfd54_n.jpg"
path1 = "flower_photos/dandelion/7355522_b66e5d3078_m.jpg"
path3 = "flower_photos/roses/394990940_7af082cf8d_n.jpg"
path4 = "flower_photos/sunflowers/6953297_8576bf4ea3.jpg"
path5 = "flower_photos/tulips/10791227_7168491604.jpg"

flower_dict = {0:'dasiy',1:'dandelion',2:'roses',3:'sunflowers',4:'tulips'}

w=100
h=100
c=3

def read_one_image(path):
    img = io.imread(path)
    img = transform.resize(img,(w,h))
    return np.asarray(img)

with tf.Session() as sess:
    data = []
    data1 = read_one_image(path1)
    data2 = read_one_image(path2)
    data3 = read_one_image(path3)
    data4 = read_one_image(path4)
    data5 = read_one_image(path5)
    data.append(data1)
    data.append(data2)
    data.append(data3)
    data.append(data4)
    data.append(data5)

    saver = tf.train.import_meta_graph('C:/Users/zhf/Desktop/flower/model/model.ckpt.meta')
    saver.restore(sess, tf.train.latest_checkpoint('C:/Users/zhf/Desktop/flower/model/'))

    graph = tf.get_default_graph()
    x = graph.get_tensor_by_name("x:0")
    feed_dict = {x : data}

    logits = graph.get_tensor_by_name("logits_eval:0")

    classification_result = sess.run(logits, feed_dict)

    # 打印出预测矩阵
    print(classification_result)

    # 打印出预测矩阵每一行最大值的索引,根据最大值的下标，对应花的类别
    print(tf.argmax(classification_result, 1).eval())

    # 根据索引通过字典对应花的分类
    output = tf.argmax(classification_result, 1).eval()

    for i in range(len(output)):
        print("第", i+1, "朵花预测:"+flower_dict[output[i]])

使用Keras训练自己的数据集——以图像多分类为例（基于卷积神经网络）