def evaluate_id_image():
max_index = []
filenames = os.listdir('test1')
for file in filenames:
print('test1/' + file)
filenames.sort(key=lambda x: int(x[:-4]))
image_array = get_one_image_file('test/' + file)
with tf.Graph().as_default():
BATCH_SIZE = 1 # 获取一张图片
N_CLASSES = 10 # 10分类
image = tf.cast(image_array, tf.float32)
image = tf.image.per_image_standardization(image)
image = tf.reshape(image, [1, 28, 28, 3]) # inference输入数据需要是4维数据，需要对image进行resize
logit = CNN.CNN_moudle(image, BATCH_SIZE, N_CLASSES, keep_prob)
logit = tf.nn.softmax(logit) # inference的softmax层没有激活函数，这里增加激活函数
# 因为只有一副图，数据量小，所以用placeholder
x = tf.placeholder(tf.float32, shape=[28, 28, 3])
#
# 训练模型路径
logs_train_dir = 'model/'
saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, str(logs_train_dir + "model.ckpt"))
prediction = sess.run(logit, feed_dict={x: image_array})
print(prediction)
# 得到概率最大的索引
index = np.argmax(prediction)
max_index.append(index)
print(max_index)
def training():
"""
##1.数据的处理
"""
# 训练图片路径
abs_path = 'train'
train_dir = os.path.abspath(abs_path)
# 输出log的位置
abs_path = 'log'
logs_train_dir = os.path.abspath(abs_path)
# 模型输出
abs_path = 'model'
train_model_dir = os.path.abspath(abs_path)
tra_list, tra_labels, val_list, val_labels = DataUtils.get_files(train_dir, 0.2)
tra_list_batch, tra_label_batch = DataUtils.get_batch(tra_list, tra_labels, IMG_W, IMG_H, BATCH_SIZE,
capacity) # 转成tensorflow 能读取的格式的数据
# val_list_batch, val_label_batch = DataUtils.get_batch(val_list, val_labels, IMG_W, IMG_H, BATCH_SIZE,
# capacity)
# print('Data Utils finished......')
# print(tra_list_batch, '\n',tra_label_batch)
# print(val_list_batch, '\n',val_label_batch)
# Tensor("batch:0", shape=(10, 28, 28, 3), dtype=float32)
# Tensor("batch:1", shape=(10, 10), dtype=int32)
# Tensor("batch_1:0", shape=(10, 28, 28, 3), dtype=float32)
# Tensor("batch_1:1", shape=(10, 10), dtype=int32)
"""
##2.网络的推理
"""
# 进行前向训练，获得回归值
prediction = CNN.CNN_moudle(tra_list_batch, BATCH_SIZE, N_CLASSES, keep_prob)
# print(prediction)
"""
##3.定义交叉熵和 要使用的梯度下降的 优化器
"""
# 二次loss
cross_entroy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=tra_label_batch,
logits=prediction))
tf.summary.scalar('loss', cross_entroy)
# Adam optimizer
train = tf.train.AdamOptimizer(learning_rate).minimize(cross_entroy)
"""
##4.定义后面要使用的变量
"""
# tf.argmax(y,1) 求标签最大的值在第几个位置 axis=1 ,表示按照行向量比较
# tf.argmax(prediction,1) 求预测最大的值在第几个位置
# 一样的 返回 tool 向量，保存起来
correct_prediction = tf.equal(tf.argmax(tra_label_batch, axis=1), tf.argmax(prediction, axis=1))
# 求准确率 True = 1.0 False = 0
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
tf.summary.scalar('accuracy', accuracy)
# 合并所有的指标
summary_op = tf.summary.merge_all()
# 新建会话
sess = tf.Session()
# 将训练日志写入到logs_train_dir的文件夹内 sess.graph:结构图
train_writer = tf.summary.FileWriter(logs_train_dir, sess.graph)
saver = tf.train.Saver() # 保存变量
# 执行训练过程，初始化变量
sess.run(tf.global_variables_initializer())
# 创建一个线程协调器，用来管理之后在Session中启动的所有线程
coord = tf.train.Coordinator()
# 启动入队的线程，一般情况下，系统有多少个核，就会启动多少个入队线程（入队具体使用多少个线程在tf.train.batch中定义）;
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
"""
进行训练：
使用 coord.should_stop()来查询是否应该终止所有线程，当文件队列（queue）中的所有文件都已经读取出列的时候，
会抛出一个 OutofRangeError 的异常，这时候就应该停止Sesson中的所有线程了;
"""
try:
for step in np.arange(MAX_STEP): # 从0 到 500 次 循环
if coord.should_stop():
break
_, tra_loss, tra_acc = sess.run([train, cross_entroy, accuracy])
# 每2步打印一次损失值和准确率
if step % 2 == 0:
print('Step %d, train loss = %.2f, train accuracy = %.4f%%' % (step, tra_loss, tra_acc * 100.0))
summary_str = sess.run(summary_op)
train_writer.add_summary(summary_str, step)
# 如果读取到文件队列末尾会抛出此异常
except tf.errors.OutOfRangeError:
print('Done training -- epoch limit reached')
finally:
coord.request_stop() # 使用coord.request_stop()来发出终止所有线程的命令
coord.join(threads) # coord.join(threads)把线程加入主线程，等待threads结束
checkpoint_path = os.path.join(train_model_dir, 'model.ckpt')
# saver.save(sess, checkpoint_path, global_step=step)
saver.save(sess, checkpoint_path)
sess.close() # 关闭会话
