tensorflow mnist
阿新 • • 發佈:2018-12-19
異常檢測測試小例子,使用mnist資料集進行正常資料集的建立,測試加入logo圖片,檢測重構誤差大於某個閾值,則判斷異常。測試logo圖片可以自己製作,某些路徑需要根據自己環境適當更改下。
import tensorflow as tf import numpy as np import tensorflow.examples.tutorials.mnist.input_data as input_data import math import scipy from scipy import ndimage mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) #------------------分割線------------------# batch_size=1 smooth=0.1 path = '/Users/changxingya/Documents/test_image' #------------------分割線------------------# #use the msra to initialize the parameters ''' input_: Tensor of input, format NHWC shape: [filter_height, filter_width, in_channel, out_channel] k_step: The size of kernel name: Name of model ''' def conv2d(input_, shape, k_step, name): with tf.variable_scope(name): msra_num = 1.0 fan_in = k_step * k_step * int(input_.get_shape()[-1]) stddev = msra_num * (math.sqrt(2. / float(fan_in))) w = tf.get_variable('w', shape, initializer= tf.truncated_normal_initializer(stddev=stddev)) b = tf.get_variable('b', [shape[-1]], initializer=tf.constant_initializer(value=0.0)) conv = tf.nn.conv2d(input_, w, strides=[1,k_step,k_step,1], padding = 'SAME') + b return conv ''' input_: Tensor of input, format NHWC output: [batch, out_height, out_width, out_channel] k_step: The size of kernel is k_step*k_step d_step: we can define the generated image size, if you define the d_step=2, we can get the double size of generated image ''' def deconv2d(input_, out_shape, k_step, d_step, name): with tf.variable_scope(name): msra_num = 1.0 fan_in = k_step * k_step * int(input_.get_shape()[-1]) stddev = msra_num * (math.sqrt(2. / float(fan_in) * float(d_step) * float(d_step))) w = tf.get_variable('w', [k_step, k_step, out_shape[-1], input_.get_shape()[-1]]) deconv = tf.nn.conv2d_transpose(input_, w, output_shape=out_shape, strides=[1, d_step, d_step, 1]) return deconv ''' input_: Tensor of input, format NHWC shape: [input_channel, out_channel] ''' def fully_contact(input_, shape, name): with tf.variable_scope(name): msra_num = 1.0 fan_in = int(input_.get_shape()[-1]) stddev = msra_num * (math.sqrt(2./float(fan_in))) w = tf.get_variable('w', shape, initializer= tf.truncated_normal_initializer(stddev=stddev)) b = tf.get_variable('b', shape[-1], initializer= tf.constant_initializer(value=0.0)) fc = tf.matmul(input_, w) + b return fc ''' define the function of leakrelu ''' def leakyrelu(x, leak=0.2): k1 = (1 + leak)*0.5 k2 = (1 - leak)*0.5 return k1 * x + k2 * tf.abs(x) ''' Restore pixel [-1, 1] to [0, 255] ''' def rescale_image(image): convert_image = (image / 1.5 + 0.5) * 255 return convert_image ''' input: The tensor of input, format NHWC size: recevie the number of images, such as size=8, we can get the 64 images, simultaneously image_path: the path to store image colorL: Ture is color image, Flase is gray image iter: record continous storage images ''' def save_image(input_, size, image_path, color, iter): h, w = input_.shape[1],input_.shape[2] convert_input = input_.reshape(batch_size, h, w) if color is True: image = np.zeros((h * size, w * size, 3)) else: image = np.zeros((h * size, w * size)) for index, img in enumerate(convert_input): i = index % size j = math.floor(index / size) if color is True: image[h*j:h*j+h, i*w:i*w+w,:] = img else: image[h*j:h*j+h, i*w:i*w+w] = img scipy.misc.toimage(rescale_image(image),cmin=0, cmax=255).save(image_path+'/tr_gt_%s.png' % (iter)) #------------------分割線------------------# def AutoEncoder(input_): with tf.variable_scope("Autoencoder", reuse = tf.AUTO_REUSE) as scope0: conv1 = conv2d(input_, [5, 5, 1, 32], 2, "conv1") conv1 = leakyrelu(conv1) conv2 = conv2d(conv1, [5, 5, 32, 64], 2, "conv2") conv2 = leakyrelu(conv2) deconv1 = deconv2d(conv2, [batch_size, 14, 14, 32], 5, 2, "deconv1") deconv1 = leakyrelu(deconv1) deconv2 = deconv2d(deconv1, [batch_size, 28, 28, 1], 5, 2, "deconv2") output = tf.tanh(deconv2) return output #------------------分割線------------------# ''' we can define the compute graphy and given the dataset to the graphy, ''' #with tf.name_scope("input"): input_image = tf.placeholder(tf.float32, [None, 28, 28, 1], 'input_image') #------------------分割線------------------# #with tf.name_scope("network"): generate_image = AutoEncoder(input_image) tf.summary.image("output_image", generate_image, 100) #------------------分割線------------------# #with tf.name_scope("loss"): Auto_loss = tf.reduce_mean(tf.reduce_sum(tf.pow(tf.subtract(generate_image, input_image), 2), 3)) tf.summary.scalar("loss", Auto_loss) #------------------分割線------------------# train_var = tf.trainable_variables() #with tf.name_scope("train"): train_loss = tf.train.AdamOptimizer(0.001, beta1=0.9).minimize(Auto_loss) init = tf.global_variables_initializer() gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.6) #------------------分割線------------------# with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options)) as sess: sess.run(init) count = 0 merged = tf.summary.merge_all() writer = tf.summary.FileWriter('/Users/changxingya/Documents/logs',sess.graph) for i in range(500): mnist_image= mnist.train.next_batch(batch_size) batch_image = mnist_image[0].reshape(batch_size, 28, 28, 1) batch_image = batch_image * 2 - 1 sess.run(train_loss, feed_dict={input_image: batch_image}) print(sess.run(Auto_loss, feed_dict={input_image: batch_image})) summary = sess.run(merged, feed_dict={input_image: batch_image}) writer.add_summary(summary, i) if i % 100 == 0: img = np.array(ndimage.imread('/Users/changxingya/Desktop/logo/'+str(count)+'.png')) img = img.reshape(1,28,28,1) count += 1 print(i) print(sess.run(Auto_loss, feed_dict={input_image: img})) output_image = sess.run(generate_image, feed_dict={input_image: img}) #result = sess.run(merged, feed_dict={input_image: batch_image}) summary= sess.run(merged, feed_dict={input_image: img}) #loss = tf.summary.scalar('loss',result) #result = sess.run(merged, feed_dict={input_image: batch_image}) writer.add_summary(summary, i) #save_image(output_image, 8, path, False, i)