网站开发方向,页面设计方案,长沙智能建站模板,网页设计制作网站教程from skimage import io, transform # skimage模块下的io transform(图像的形变与缩放)模块
import glob # glob 文件通配符模块
import os # os 处理文件和目录的模块
import tensorflow as tf
import numpy as np # 多维数据处理模块
import time
import matplotlib.pypl…from skimage import io, transform # skimage模块下的io transform(图像的形变与缩放)模块
import glob # glob 文件通配符模块
import os # os 处理文件和目录的模块
import tensorflow as tf
import numpy as np # 多维数据处理模块
import time
import matplotlib.pyplot as plt
# 数据集地址
#path ./flower_photos/
path ../dataset/train/
# 模型保存地址
model_path ./cnn1_model.ckpt# 将所有的图片resize成100*100
w 100
h 100
c 3# 读取图片数据处理
def read_img(path):# os.listdir(path) 返回path指定的文件夹包含的文件或文件夹的名字的列表# os.path.isdir(path)判断path是否是目录# b [xx for x in list1 if xx15 ] 列表生成式,循环list1当if为真时将xx加入列表bcate [path x for x in os.listdir(path) if os.path.isdir(path x)]imgs []labels []print(开始读入图片和标签。。。。)for idx, folder in enumerate(cate):# glob.glob(s*.py) 从目录通配符搜索中生成文件列表for im in glob.glob(folder /*.png):# 输出读取的图片的名称#print(reading the images:%s % (im))# io.imread(im)读取单张RGB图片 skimage.io.imread(fname,as_greyTrue)读取单张灰度图片# 读取的图片img io.imread(im)# skimage.transform.resize(image, output_shape)改变图片的尺寸img transform.resize(img, (w, h))# 将读取的图片数据加载到imgs[]列表中imgs.append(img)# 将图片的label加载到labels[]中与上方的imgs索引对应labels.append(idx)# 将读取的图片和labels信息转化为numpy结构的ndarr(N维数组对象矩阵)数据信息print(读入图片和标签完毕。。。。)return np.asarray(imgs, np.float32), np.asarray(labels, np.int32)# 调用读取图片的函数得到图片和labels的数据集
data, label read_img(path)# 打乱顺序
# 读取data矩阵的第一维数图片的个数
num_example data.shape[0]
# 产生一个num_example范围步长为1的序列
arr np.arange(num_example)
# 调用函数打乱顺序
np.random.shuffle(arr)
# 按照打乱的顺序重新排序
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----------------------
# 本程序cnn网络模型共有7层前三层为卷积层后三层为全连接层前三层中每层包含卷积、激活、池化层
# 占位符设置输入参数的大小和格式
x tf.placeholder(tf.float32, shape[None, w, h, c], namex)
y_ tf.placeholder(tf.int32, shape[None, ], namey_)def inference(input_tensor, train, regularizer):# -----------------------第一层----------------------------with tf.variable_scope(layer1-conv1):# 初始化权重conv1_weights为可保存变量大小为5x5,3个通道RGB数量为32个conv1_weights tf.get_variable(weight, [5, 5, 3, 32],initializertf.truncated_normal_initializer(stddev0.1))# 初始化偏置conv1_biases数量为32个conv1_biases tf.get_variable(bias, [32], initializertf.constant_initializer(0.0))# 卷积计算tf.nn.conv2d为tensorflow自带2维卷积函数input_tensor为输入数据# conv1_weights为权重strides[1, 1, 1, 1]表示左右上下滑动步长为1paddingSAME表示输入和输出大小一样即补0conv1 tf.nn.conv2d(input_tensor, conv1_weights, strides[1, 1, 1, 1], paddingSAME)# 激励计算调用tensorflow的relu函数relu1 tf.nn.relu(tf.nn.bias_add(conv1, conv1_biases))with tf.name_scope(layer2-pool1):# 池化计算调用tensorflow的max_pool函数strides[1,2,2,1]表示池化边界2个对一个生成paddingVALID表示不操作。pool1 tf.nn.max_pool(relu1, ksize[1, 2, 2, 1], strides[1, 2, 2, 1], paddingVALID)# -----------------------第二层----------------------------with tf.variable_scope(layer3-conv2):# 同上不过参数的有变化根据卷积计算和通道数量的变化设置对应的参数conv2_weights tf.get_variable(weight, [5, 5, 32, 64],initializertf.truncated_normal_initializer(stddev0.1))conv2_biases tf.get_variable(bias, [64], initializertf.constant_initializer(0.0))conv2 tf.nn.conv2d(pool1, conv2_weights, strides[1, 1, 1, 1], paddingSAME)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], paddingVALID)# -----------------------第三层----------------------------# 同上不过参数的有变化根据卷积计算和通道数量的变化设置对应的参数with tf.variable_scope(layer5-conv3):conv3_weights tf.get_variable(weight, [3, 3, 64, 128],initializertf.truncated_normal_initializer(stddev0.1))conv3_biases tf.get_variable(bias, [128], initializertf.constant_initializer(0.0))conv3 tf.nn.conv2d(pool2, conv3_weights, strides[1, 1, 1, 1], paddingSAME)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], paddingVALID)# -----------------------第四层----------------------------# 同上不过参数的有变化根据卷积计算和通道数量的变化设置对应的参数with tf.variable_scope(layer7-conv4):conv4_weights tf.get_variable(weight, [3, 3, 128, 128],initializertf.truncated_normal_initializer(stddev0.1))conv4_biases tf.get_variable(bias, [128], initializertf.constant_initializer(0.0))conv4 tf.nn.conv2d(pool3, conv4_weights, strides[1, 1, 1, 1], paddingSAME)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], paddingVALID)nodes 6 * 6 * 128reshaped tf.reshape(pool4, [-1, nodes])# 使用变形函数转化结构# -----------------------第五层---------------------------with tf.variable_scope(layer9-fc1):# 初始化全连接层的参数隐含节点为1024个fc1_weights tf.get_variable(weight, [nodes, 1024],initializertf.truncated_normal_initializer(stddev0.1))if regularizer ! None: tf.add_to_collection(losses, regularizer(fc1_weights)) # 正则化矩阵fc1_biases tf.get_variable(bias, [1024], initializertf.constant_initializer(0.1))# 使用relu函数作为激活函数fc1 tf.nn.relu(tf.matmul(reshaped, fc1_weights) fc1_biases)# 采用dropout层减少过拟合和欠拟合的程度保存模型最好的预测效率if train: fc1 tf.nn.dropout(fc1, 0.5)# -----------------------第六层----------------------------with tf.variable_scope(layer10-fc2):# 同上不过参数的有变化根据卷积计算和通道数量的变化设置对应的参数fc2_weights tf.get_variable(weight, [1024, 512],initializertf.truncated_normal_initializer(stddev0.1))if regularizer ! None: tf.add_to_collection(losses, regularizer(fc2_weights))fc2_biases tf.get_variable(bias, [512], initializertf.constant_initializer(0.1))fc2 tf.nn.relu(tf.matmul(fc1, fc2_weights) fc2_biases)if train: fc2 tf.nn.dropout(fc2, 0.5)# -----------------------第七层----------------------------with tf.variable_scope(layer11-fc3):# 同上不过参数的有变化根据卷积计算和通道数量的变化设置对应的参数fc3_weights tf.get_variable(weight, [512, 5],initializertf.truncated_normal_initializer(stddev0.1))if regularizer ! None: tf.add_to_collection(losses, regularizer(fc3_weights))fc3_biases tf.get_variable(bias, [5], initializertf.constant_initializer(0.1))logit tf.matmul(fc2, fc3_weights) fc3_biases # matmul矩阵相乘# 返回最后的计算结果return logit# ---------------------------网络结束---------------------------
# 设置正则化参数为0.0001
regularizer tf.contrib.layers.l2_regularizer(0.0001)
# 将上述构建网络结构引入
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(value1, dtypetf.float32)
logits_eval tf.multiply(logits, b, namelogits_eval) # b为1# 设置损失函数作为模型训练优化的参考标准loss越小模型越优
loss tf.nn.sparse_softmax_cross_entropy_with_logits(logitslogits, labelsy_)
# 设置整体学习率为α为0.001
train_op tf.train.AdamOptimizer(learning_rate0.001).minimize(loss)
# 设置预测精度
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(inputsNone, targetsNone, batch_sizeNone, shuffleFalse):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): #range(start,end,step)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# 迭代次数
n_epoch 30
fig_loss np.zeros([n_epoch])
fig_acc1 np.zeros([n_epoch])
fig_acc2 np.zeros([n_epoch])
# 每次迭代输入的图片数据
batch_size 64
saver tf.train.Saver(max_to_keep1) # 可以指定保存的模型个数利用max_to_keep4则最终会保存4个模型
with tf.Session() as sess:# 初始化全局参数sess.run(tf.global_variables_initializer())# 开始迭代训练调用的都是前面设置好的函数或变量for epoch in range(n_epoch):start_time time.time()# training#训练集train_loss, train_acc, n_batch 0, 0, 0for x_train_a, y_train_a in minibatches(x_train, y_train, batch_size, shuffleTrue):_, err, ac sess.run([train_op, loss, acc], feed_dict{x: x_train_a, y_: y_train_a})train_loss errtrain_acc acn_batch 1if n_batch%200:# print(Epoch:%d After %d batch_size train loss % (n_epoch,n_batch))# print(err)print(Epoch:%d After %d batch_size average train loss: %f % (epoch, n_batch, np.sum(train_loss) / n_batch))# print(Epoch:%d After %d batch_size train acc %f % (epoch, n_batch,ac))print(Epoch:%d After %d batch_size average train acc: %f % (epoch, n_batch, np.sum(train_acc) / n_batch))#Epoch: 9 After 45 batch_size average train loss: 2.750402 Epoch: 9#After 45 batch_size average train acc: 0.993403fig_loss[epoch] np.sum(train_loss) / n_batchfig_acc1[epoch] np.sum(train_acc) / n_batch#validation#验证集val_loss, val_acc, n_batch 0, 0, 0for x_val_a, y_val_a in minibatches(x_val, y_val, batch_size, shuffleFalse):err, ac sess.run([loss, acc], feed_dict{x: x_val_a, y_: y_val_a})val_loss errval_acc acn_batch 1print(validation loss: %f % (np.sum(val_loss) / n_batch))print(validation acc: %f % (np.sum(val_acc) / n_batch))fig_acc2[epoch] np.sum(val_acc) / n_batch#保存模型及模型参数if epoch % 2 0:saver.save(sess, model_path, global_stepepoch)# 训练loss图
fig, ax1 plt.subplots()
lns1 ax1.plot(np.arange(n_epoch), fig_loss, labelLoss)
ax1.set_xlabel(iteration)
ax1.set_ylabel(training loss)# 训练和验证两种准确率曲线图放在一张图中
fig2, ax2 plt.subplots()
ax3 ax2.twinx()#由ax2图生成ax3图
lns2 ax2.plot(np.arange(n_epoch), fig_acc1, labelLoss)
lns3 ax3.plot(np.arange(n_epoch), fig_acc2, labelLoss)ax2.set_xlabel(iteration)
ax2.set_ylabel(training acc)
ax3.set_ylabel(val acc)# 合并图例
lns lns3 lns2
labels [train acc, val acc]
plt.legend(lns, labels, loc7)plt.show()
