模式识别（九）神经网络（RNN）识别MNIST和细胞数据集

MNIST数据集

from tensorflow.contrib.layers import fully_connected
import tensorflow as tf

n_steps = 28
n_inputs = 28
n_neurons = 150
n_outputs = 10
learning_rate = 0.001
# 占位符
X = tf.placeholder(tf.float32, [None, n_steps, n_inputs])
y = tf.placeholder(tf.int32, [None])
# RNN单元
basic_cell = tf.contrib.rnn.BasicRNNCell(num_units=n_neurons)  # RNN基本单元，num_units为单元个数
outputs, states = tf.nn.dynamic_rnn(basic_cell, X, dtype=tf.float32)
# 全连接
logits = fully_connected(states, n_outputs, activation_fn=None)
# 损失函数及优化器
xentropy = tf.nn.sparse_softmax_cross_entropy_with_logits(labels=y, logits=logits)
loss = tf.reduce_mean(xentropy)
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss)
# 准确率
correct = tf.nn.in_top_k(logits, y, 1)
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
# 初始化
init = tf.global_variables_initializer()

from tensorflow.examples.tutorials.mnist import input_data

mnist = input_data.read_data_sets("/tmp/data/")
X_test = mnist.test.images.reshape((-1, n_steps, n_inputs))
y_test = mnist.test.labels
# 训练，与前面DNN类似，不在叙述了
n_epochs = 100
batch_size = 150
with tf.Session() as sess:
    init.run()
    for epoch in range(n_epochs):
        for iteration in range(mnist.train.num_examples // batch_size):
            X_batch, y_batch = mnist.train.next_batch(batch_size)
            X_batch = X_batch.reshape((-1, n_steps, n_inputs))
            sess.run(train_op, feed_dict={X: X_batch, y: y_batch})
        acc_train = accuracy.eval(feed_dict={X: X_batch, y: y_batch})
        acc_test = accuracy.eval(feed_dict={X: X_test, y: y_test})
        print(epoch, "Train accuracy:", acc_train, "Test accuracy:", acc_test)

细胞数据集

import os
import time
import cv2 as cv
import numpy as np
import tensorflow as tf
from keras.models import Sequential
from keras.layers import (Dense, Dropout, SimpleRNN, Embedding)
n_steps = 28
n_inputs = 28
n_neurons = 150
n_outputs = 10
learning_rate = 0.001
# ——————————————————————————————————————————————function————————————————————————————————————————————————————
def set_gpus(gpu_index):
    if type(gpu_index) == list:
        gpu_index = ','.join(str(_) for _ in gpu_index)
    if type(gpu_index) ==int:
        gpu_index = str(gpu_index)
    os.environ["CUDA_VISIBLE_DEVICES"] = gpu_index
def get_path(filepath):
    filelist = []
    filedirs = []
    for root, _, files in os.walk(filepath, topdown = False):
        for name in files:
            filelist.append(name)
            filedirs.append(os.path.join(root, name))
    return filelist, filedirs
def read_file(path,filelist, filedirs, Use_smote = False):
    dic ={'DCIS':0, 'IDC':0, 'Muc':0, 'ILC':0, 'MC':0,
          'normal':1,
          'TIL':2}
    datalist = []
    claslist = []
    for path in filedirs:
        img = cv.imread(path, 0) #单通道读入灰度图像
        img = img.flatten() / 255
        name = filelist[filedirs.index(path)]
        clas = name.split('_')[0] #保留_前面的字符串即：'DCIS'...
        datalist.append(img)
        claslist.append(dic[clas])
    claslist = np.array(claslist)
    datalist = np.array(datalist)
    return datalist, claslist
def RNNnetwork(datalist, claslist):
    # 输入图片形状(14536, 51, 51)
    model = Sequential()
    model.add(Embedding(input_dim=1, output_dim=1))
    model.add(Dropout(0.5))
    model.add(SimpleRNN(16))
    model.add(Dense(256, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(1, activation='sigmoid'))
    model.summary()
    model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])
    model.fit(datalist, claslist, epochs=10, batch_size=100)

    sess = tf.Session()
    init = tf.global_variables_initializer()
    sess.run(init)

    return model.evaluate(datalist, claslist, verbose=2)

# ————————————————————————————————————————————————main——————————————————————————————————————————————————————
set_gpus(0)
time_start = time.time()
path = 'C:\\Users\\Administrator\\Desktop\\Tensorflow\\cells\\train'
filepath, filedirs = get_path(path)
datalist, claslist = read_file(path, filepath, filedirs, Use_smote = False)
print('读入数据完毕，开始训练')
results = RNNnetwork(datalist, claslist)
print('RNN的loss和正确率分别为：', results)
time_end = time.time()
print('耗时：', time_end-time_start, 's')