上一次修改时间：2018-08-21 22:25:14
多任务学习识别验证码
1.验证码图片生成
# 验证码生成库
from captcha.image import ImageCaptcha # pip install captcha
import numpy as np
from PIL import Image
import random
import sys
number = ['0','1','2','3','4','5','6','7','8','9']
# alphabet = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z']
# ALPHABET = ['A','B','C','D','E','F','G','H','I','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z']
#验证码字符集加上小写字母示例：char_set=number+alphabet
def random_captcha_text(char_set=number, captcha_size=4):
# 验证码列表
captcha_text = []
for i in range(captcha_size):
#随机选择
c = random.choice(char_set)
#加入验证码列表
captcha_text.append(c)
return captcha_text
# 生成字符对应的验证码
def gen_captcha_text_and_image():
image = ImageCaptcha()
#获得随机生成的验证码
captcha_text = random_captcha_text()
#把验证码列表转为字符串
captcha_text = ''.join(captcha_text)
#生成验证码
captcha = image.generate(captcha_text)
image.write(captcha_text, 'captcha/images/' + captcha_text + '.jpg') # 写到文件
#数量少于10000，因为重名
num = 10000
if __name__ == '__main__':
for i in range(num):
gen_captcha_text_and_image()
sys.stdout.write('\r>> Creating image %d/%d' % (i+1, num))
sys.stdout.flush()
sys.stdout.write('\n')
sys.stdout.flush()
print("生成完毕")
生成的验证图片如图：
2.验证码图片生成tfrecored文件
tfrecored是tensorflow用来存储数据的格式；
import tensorflow as tf import os import random import math import sys from PIL import Image import numpy as np
#验证集数量
_NUM_TEST = 500
#随机种子
_RANDOM_SEED = 0
#数据集路径
DATASET_DIR = "D:/pikaqiu/captcha/images/"
#tfrecord文件存放路径
TFRECORD_DIR = "D:/pikaqiu/captcha/"
#判断tfrecord文件是否存在
def _dataset_exists(dataset_dir):
for split_name in ['train', 'test']:
output_filename = os.path.join(dataset_dir,split_name + '.tfrecords')
if not tf.gfile.Exists(output_filename):
return False
return True
#获取所有验证码图片
def _get_filenames_and_classes(dataset_dir):
photo_filenames = []
for filename in os.listdir(dataset_dir):
#获取文件路径
path = os.path.join(dataset_dir, filename)
photo_filenames.append(path)
return photo_filenames
def int64_feature(values):
if not isinstance(values, (tuple, list)):
values = [values]
return tf.train.Feature(int64_list=tf.train.Int64List(value=values))
def bytes_feature(values):
return tf.train.Feature(bytes_list=tf.train.BytesList(value=[values]))
def image_to_tfexample(image_data, label0, label1, label2, label3):
#Abstract base class for protocol messages.
return tf.train.Example(features=tf.train.Features(feature={
'image': bytes_feature(image_data),
'label0': int64_feature(label0),
'label1': int64_feature(label1),
'label2': int64_feature(label2),
'label3': int64_feature(label3),
}))
#把数据转为TFRecord格式
def _convert_dataset(split_name, filenames, dataset_dir):
assert split_name in ['train', 'test']
with tf.Session() as sess:
#定义tfrecord文件的路径+名字
output_filename = os.path.join(TFRECORD_DIR,split_name + '.tfrecords')
with tf.python_io.TFRecordWriter(output_filename) as tfrecord_writer:
for i,filename in enumerate(filenames):
try:
sys.stdout.write('\r>> Converting image %d/%d' % (i+1, len(filenames)))
sys.stdout.flush()
#读取图片
image_data = Image.open(filename)
#根据模型的结构resize,训练时的模型要求的输入为224*224
image_data = image_data.resize((224, 224))
#灰度化，验证码识别时，颜色不对结果产生影响，因此需要做灰度化
#灰度化后需要计算的通道只有一个，而彩色时有三个，相较下，计算量会小很多
image_data = np.array(image_data.convert('L'))
#将图片转化为bytes
image_data = image_data.tobytes()
#获取label
labels = filename.split('/')[-1][0:4]
num_labels = []
for j in range(4):
num_labels.append(int(labels[j]))
#生成protocol数据类型
example = image_to_tfexample(image_data, num_labels[0], num_labels[1], num_labels[2], num_labels[3])
tfrecord_writer.write(example.SerializeToString())
except IOError as e:
print('Could not read:',filename)
print('Error:',e)
print('Skip it\n')
sys.stdout.write('\n')
sys.stdout.flush()
#判断tfrecord文件是否存在
if _dataset_exists(TFRECORD_DIR):
print('tfcecord文件已存在')
else:
#获得所有图片
photo_filenames = _get_filenames_and_classes(DATASET_DIR)
#把数据切分为训练集和测试集,并打乱
random.seed(_RANDOM_SEED)
random.shuffle(photo_filenames)
training_filenames = photo_filenames[_NUM_TEST:]
testing_filenames = photo_filenames[:_NUM_TEST]
#数据转换
_convert_dataset('train', training_filenames, DATASET_DIR)
_convert_dataset('test', testing_filenames, DATASET_DIR)
print('生成tfcecord文件')
生成的文件如图：
3.模型训练
使用的模型为alexnet_v2
import os import tensorflow as tf from PIL import Image #ntes是从github上下载下来的源码里的文件夹 #示例中使用的alexnet_v2有部分修改 from nets import nets_factory import numpy as np
# 不同字符数量
CHAR_SET_LEN = 10
# 图片高度
IMAGE_HEIGHT = 60
# 图片宽度
IMAGE_WIDTH = 160
# 批次
BATCH_SIZE = 25
# tfrecord文件存放路径
TFRECORD_FILE = "D:/pikaqiu/captcha/train.tfrecords"
# placeholder
x = tf.placeholder(tf.float32, [None, 224, 224])
y0 = tf.placeholder(tf.float32, [None])
y1 = tf.placeholder(tf.float32, [None])
y2 = tf.placeholder(tf.float32, [None])
y3 = tf.placeholder(tf.float32, [None])
# 学习率
lr = tf.Variable(0.003, dtype=tf.float32)
# 从tfrecord读出数据
def read_and_decode(filename):
# 根据文件名生成一个队列
filename_queue = tf.train.string_input_producer([filename])
reader = tf.TFRecordReader()
# 返回文件名和文件
_, serialized_example = reader.read(filename_queue)
features = tf.parse_single_example(serialized_example,
features={
'image' : tf.FixedLenFeature([], tf.string),
'label0': tf.FixedLenFeature([], tf.int64),
'label1': tf.FixedLenFeature([], tf.int64),
'label2': tf.FixedLenFeature([], tf.int64),
'label3': tf.FixedLenFeature([], tf.int64),
})
# 获取图片数据
image = tf.decode_raw(features['image'], tf.uint8)
# tf.train.shuffle_batch必须确定shape
image = tf.reshape(image, [224, 224])
# 图片预处理
image = tf.cast(image, tf.float32) / 255.0
image = tf.subtract(image, 0.5)
image = tf.multiply(image, 2.0)
# 获取label
label0 = tf.cast(features['label0'], tf.int32)
label1 = tf.cast(features['label1'], tf.int32)
label2 = tf.cast(features['label2'], tf.int32)
label3 = tf.cast(features['label3'], tf.int32)
return image, label0, label1, label2, label3
# In[3]:
# 获取图片数据和标签
image, label0, label1, label2, label3 = read_and_decode(TFRECORD_FILE)
#使用shuffle_batch可以随机打乱
image_batch, label_batch0, label_batch1, label_batch2, label_batch3 = tf.train.shuffle_batch(
[image, label0, label1, label2, label3], batch_size = BATCH_SIZE,
capacity = 50000, min_after_dequeue=10000, num_threads=1)
#定义网络结构
train_network_fn = nets_factory.get_network_fn(
'alexnet_v2',
num_classes=CHAR_SET_LEN,
weight_decay=0.0005,
is_training=True)
with tf.Session() as sess:
# inputs: a tensor of size [batch_size, height, width, channels]
X = tf.reshape(x, [BATCH_SIZE, 224, 224, 1])
# 数据输入网络得到输出值
logits0,logits1,logits2,logits3,end_points = train_network_fn(X)
# 把标签转成one_hot的形式
one_hot_labels0 = tf.one_hot(indices=tf.cast(y0, tf.int32), depth=CHAR_SET_LEN)
one_hot_labels1 = tf.one_hot(indices=tf.cast(y1, tf.int32), depth=CHAR_SET_LEN)
one_hot_labels2 = tf.one_hot(indices=tf.cast(y2, tf.int32), depth=CHAR_SET_LEN)
one_hot_labels3 = tf.one_hot(indices=tf.cast(y3, tf.int32), depth=CHAR_SET_LEN)
# 计算loss
loss0 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits0,labels=one_hot_labels0))
loss1 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits1,labels=one_hot_labels1))
loss2 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits2,labels=one_hot_labels2))
loss3 = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits3,labels=one_hot_labels3))
# 计算总的loss
total_loss = (loss0+loss1+loss2+loss3)/4.0
# 优化total_loss
optimizer = tf.train.AdamOptimizer(learning_rate=lr).minimize(total_loss)
# 计算准确率
correct_prediction0 = tf.equal(tf.argmax(one_hot_labels0,1),tf.argmax(logits0,1))
accuracy0 = tf.reduce_mean(tf.cast(correct_prediction0,tf.float32))
correct_prediction1 = tf.equal(tf.argmax(one_hot_labels1,1),tf.argmax(logits1,1))
accuracy1 = tf.reduce_mean(tf.cast(correct_prediction1,tf.float32))
correct_prediction2 = tf.equal(tf.argmax(one_hot_labels2,1),tf.argmax(logits2,1))
accuracy2 = tf.reduce_mean(tf.cast(correct_prediction2,tf.float32))
correct_prediction3 = tf.equal(tf.argmax(one_hot_labels3,1),tf.argmax(logits3,1))
accuracy3 = tf.reduce_mean(tf.cast(correct_prediction3,tf.float32))
# 用于保存模型
saver = tf.train.Saver()
# 初始化
sess.run(tf.global_variables_initializer())
# 创建一个协调器，管理线程
coord = tf.train.Coordinator()
# 启动QueueRunner, 此时文件名队列已经进队
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
for i in range(6001):
# 获取一个批次的数据和标签
b_image, b_label0, b_label1 ,b_label2 ,b_label3 = sess.run([image_batch, label_batch0, label_batch1, label_batch2, label_batch3])
# 优化模型
sess.run(optimizer, feed_dict={x: b_image, y0:b_label0, y1: b_label1, y2: b_label2, y3: b_label3})
# 每迭代20次计算一次loss和准确率
if i % 20 == 0:
# 每迭代2000次降低一次学习率
if i%2000 == 0:
sess.run(tf.assign(lr, lr/3))
acc0,acc1,acc2,acc3,loss_ = sess.run([accuracy0,accuracy1,accuracy2,accuracy3,total_loss],feed_dict={x: b_image,
y0: b_label0,
y1: b_label1,
y2: b_label2,
y3: b_label3})
learning_rate = sess.run(lr)
print ("Iter:%d Loss:%.3f Accuracy:%.2f,%.2f,%.2f,%.2f Learning_rate:%.4f" % (i,loss_,acc0,acc1,acc2,acc3,learning_rate))
# 保存模型
# if acc0 > 0.90 and acc1 > 0.90 and acc2 > 0.90 and acc3 > 0.90:
if i==6000:
saver.save(sess, "./captcha/models/crack_captcha.model", global_step=i)
break
# 通知其他线程关闭
coord.request_stop()
# 其他所有线程关闭之后，这一函数才能返回
coord.join(threads)运行结果如下，经过6000次迭代后，准确率差不多能达到100%：
Iter:0 Loss:1713.653 Accuracy:0.24,0.24,0.16,0.32 Learning_rate:0.0010 Iter:20 Loss:2.303 Accuracy:0.16,0.00,0.08,0.08 Learning_rate:0.0010 Iter:40 Loss:2.308 Accuracy:0.16,0.04,0.08,0.12 Learning_rate:0.0010 Iter:60 Loss:2.296 Accuracy:0.16,0.04,0.28,0.08 Learning_rate:0.0010 Iter:80 Loss:2.298 Accuracy:0.20,0.00,0.20,0.08 Learning_rate:0.0010 Iter:100 Loss:2.303 Accuracy:0.16,0.08,0.04,0.08 Learning_rate:0.0010 Iter:120 Loss:2.297 Accuracy:0.12,0.08,0.08,0.04 Learning_rate:0.0010 Iter:140 Loss:2.300 Accuracy:0.08,0.12,0.24,0.16 Learning_rate:0.0010 Iter:160 Loss:2.297 Accuracy:0.12,0.08,0.16,0.20 Learning_rate:0.0010 Iter:180 Loss:2.305 Accuracy:0.08,0.08,0.08,0.16 Learning_rate:0.0010 Iter:200 Loss:2.301 Accuracy:0.20,0.04,0.16,0.12 Learning_rate:0.0010 Iter:220 Loss:2.295 Accuracy:0.04,0.12,0.04,0.12 Learning_rate:0.0010 Iter:240 Loss:2.311 Accuracy:0.16,0.16,0.08,0.08 Learning_rate:0.0010 Iter:260 Loss:2.297 Accuracy:0.08,0.00,0.20,0.08 Learning_rate:0.0010 Iter:280 Loss:2.298 Accuracy:0.20,0.12,0.12,0.16 Learning_rate:0.0010 Iter:300 Loss:2.307 Accuracy:0.16,0.16,0.12,0.00 Learning_rate:0.0010 Iter:320 Loss:2.276 Accuracy:0.04,0.04,0.20,0.20 Learning_rate:0.0010 Iter:340 Loss:2.276 Accuracy:0.12,0.12,0.16,0.16 Learning_rate:0.0010 Iter:360 Loss:2.301 Accuracy:0.16,0.24,0.24,0.08 Learning_rate:0.0010 Iter:380 Loss:2.275 Accuracy:0.08,0.12,0.20,0.20 Learning_rate:0.0010 Iter:400 Loss:2.184 Accuracy:0.28,0.12,0.12,0.16 Learning_rate:0.0010 Iter:420 Loss:2.168 Accuracy:0.28,0.16,0.20,0.20 Learning_rate:0.0010 Iter:440 Loss:2.221 Accuracy:0.00,0.12,0.24,0.24 Learning_rate:0.0010 Iter:460 Loss:2.197 Accuracy:0.08,0.24,0.16,0.20 Learning_rate:0.0010 Iter:480 Loss:2.106 Accuracy:0.24,0.12,0.12,0.20 Learning_rate:0.0010 Iter:500 Loss:2.022 Accuracy:0.32,0.20,0.16,0.28 Learning_rate:0.0010 Iter:520 Loss:1.874 Accuracy:0.32,0.40,0.28,0.20 Learning_rate:0.0010 Iter:540 Loss:2.048 Accuracy:0.52,0.12,0.16,0.28 Learning_rate:0.0010 Iter:560 Loss:1.781 Accuracy:0.48,0.24,0.20,0.20 Learning_rate:0.0010 Iter:580 Loss:1.943 Accuracy:0.48,0.16,0.28,0.32 Learning_rate:0.0010 Iter:600 Loss:1.779 Accuracy:0.48,0.36,0.28,0.44 Learning_rate:0.0010 Iter:620 Loss:1.598 Accuracy:0.56,0.40,0.24,0.52 Learning_rate:0.0010 Iter:640 Loss:1.539 Accuracy:0.60,0.32,0.36,0.48 Learning_rate:0.0010 Iter:660 Loss:1.414 Accuracy:0.80,0.44,0.20,0.52 Learning_rate:0.0010 Iter:680 Loss:1.358 Accuracy:0.76,0.32,0.36,0.44 Learning_rate:0.0010 Iter:700 Loss:1.120 Accuracy:0.80,0.76,0.56,0.48 Learning_rate:0.0010 Iter:720 Loss:1.265 Accuracy:0.60,0.56,0.32,0.72 Learning_rate:0.0010 Iter:740 Loss:1.091 Accuracy:0.84,0.52,0.48,0.52 Learning_rate:0.0010 Iter:760 Loss:1.187 Accuracy:0.60,0.44,0.48,0.56 Learning_rate:0.0010 Iter:780 Loss:1.161 Accuracy:0.68,0.52,0.32,0.68 Learning_rate:0.0010 Iter:800 Loss:0.988 Accuracy:0.76,0.48,0.52,0.64 Learning_rate:0.0010 Iter:820 Loss:1.068 Accuracy:0.84,0.56,0.52,0.64 Learning_rate:0.0010 Iter:840 Loss:0.906 Accuracy:0.80,0.48,0.72,0.80 Learning_rate:0.0010 Iter:860 Loss:0.928 Accuracy:0.92,0.44,0.60,0.68 Learning_rate:0.0010 Iter:880 Loss:1.193 Accuracy:0.52,0.64,0.52,0.56 Learning_rate:0.0010 Iter:900 Loss:0.888 Accuracy:0.68,0.64,0.44,0.64 Learning_rate:0.0010 Iter:920 Loss:0.772 Accuracy:0.80,0.80,0.60,0.80 Learning_rate:0.0010 Iter:940 Loss:0.862 Accuracy:0.84,0.68,0.56,0.76 Learning_rate:0.0010 Iter:960 Loss:0.991 Accuracy:0.88,0.56,0.52,0.64 Learning_rate:0.0010 Iter:980 Loss:0.874 Accuracy:0.80,0.76,0.64,0.68 Learning_rate:0.0010 Iter:1000 Loss:0.658 Accuracy:0.92,0.72,0.56,0.80 Learning_rate:0.0010 Iter:1020 Loss:0.808 Accuracy:0.80,0.76,0.60,0.68 Learning_rate:0.0010 Iter:1040 Loss:0.635 Accuracy:0.88,0.72,0.68,0.80 Learning_rate:0.0010 Iter:1060 Loss:0.787 Accuracy:0.72,0.76,0.80,0.64 Learning_rate:0.0010 Iter:1080 Loss:0.608 Accuracy:0.84,0.76,0.76,0.80 Learning_rate:0.0010 Iter:1100 Loss:0.556 Accuracy:0.88,0.80,0.72,0.72 Learning_rate:0.0010 Iter:1120 Loss:0.740 Accuracy:0.92,0.68,0.64,0.64 Learning_rate:0.0010 Iter:1140 Loss:0.548 Accuracy:0.92,0.76,0.64,0.84 Learning_rate:0.0010 Iter:1160 Loss:0.598 Accuracy:0.96,0.60,0.72,0.96 Learning_rate:0.0010 Iter:1180 Loss:0.661 Accuracy:0.84,0.68,0.76,0.76 Learning_rate:0.0010 Iter:1200 Loss:0.789 Accuracy:0.84,0.52,0.64,0.72 Learning_rate:0.0010 Iter:1220 Loss:0.664 Accuracy:0.68,0.64,0.80,0.84 Learning_rate:0.0010 Iter:1240 Loss:0.626 Accuracy:1.00,0.80,0.64,0.88 Learning_rate:0.0010 Iter:1260 Loss:0.657 Accuracy:0.96,0.80,0.76,0.60 Learning_rate:0.0010 Iter:1280 Loss:0.488 Accuracy:0.92,0.80,0.72,0.76 Learning_rate:0.0010 Iter:1300 Loss:0.465 Accuracy:0.96,0.68,0.80,0.76 Learning_rate:0.0010 Iter:1320 Loss:0.457 Accuracy:0.92,0.84,0.72,0.84 Learning_rate:0.0010 Iter:1340 Loss:0.502 Accuracy:0.80,0.76,0.88,0.80 Learning_rate:0.0010 Iter:1360 Loss:0.433 Accuracy:0.84,0.76,0.92,0.92 Learning_rate:0.0010 Iter:1380 Loss:0.468 Accuracy:0.92,0.80,0.88,0.84 Learning_rate:0.0010 Iter:1400 Loss:0.400 Accuracy:0.92,0.84,0.88,0.80 Learning_rate:0.0010 Iter:1420 Loss:0.395 Accuracy:0.84,0.76,0.76,0.92 Learning_rate:0.0010 Iter:1440 Loss:0.413 Accuracy:0.96,0.80,0.84,0.76 Learning_rate:0.0010 Iter:1460 Loss:0.345 Accuracy:0.92,0.96,0.76,0.88 Learning_rate:0.0010 Iter:1480 Loss:0.462 Accuracy:0.80,0.88,0.80,0.92 Learning_rate:0.0010 Iter:1500 Loss:0.432 Accuracy:0.92,0.80,0.76,0.96 Learning_rate:0.0010 Iter:1520 Loss:0.217 Accuracy:1.00,0.84,1.00,0.88 Learning_rate:0.0010 Iter:1540 Loss:0.524 Accuracy:0.92,0.84,0.68,0.76 Learning_rate:0.0010 Iter:1560 Loss:0.453 Accuracy:0.84,0.76,0.76,0.84 Learning_rate:0.0010 Iter:1580 Loss:0.358 Accuracy:1.00,0.84,0.76,0.76 Learning_rate:0.0010 Iter:1600 Loss:0.268 Accuracy:0.96,0.96,0.76,0.92 Learning_rate:0.0010 Iter:1620 Loss:0.275 Accuracy:0.88,0.96,0.88,0.92 Learning_rate:0.0010 Iter:1640 Loss:0.479 Accuracy:0.80,0.80,0.92,0.80 Learning_rate:0.0010 Iter:1660 Loss:0.291 Accuracy:0.92,0.92,0.84,0.88 Learning_rate:0.0010 Iter:1680 Loss:0.517 Accuracy:0.88,0.80,0.84,0.80 Learning_rate:0.0010 Iter:1700 Loss:0.280 Accuracy:0.80,0.88,0.88,0.88 Learning_rate:0.0010 Iter:1720 Loss:0.432 Accuracy:0.80,0.80,0.76,0.88 Learning_rate:0.0010 Iter:1740 Loss:0.376 Accuracy:1.00,0.96,0.76,0.72 Learning_rate:0.0010 Iter:1760 Loss:0.319 Accuracy:0.92,0.88,0.72,0.92 Learning_rate:0.0010 Iter:1780 Loss:0.329 Accuracy:0.92,0.88,0.88,0.88 Learning_rate:0.0010 Iter:1800 Loss:0.249 Accuracy:0.96,1.00,0.88,0.92 Learning_rate:0.0010 Iter:1820 Loss:0.351 Accuracy:0.96,0.88,0.80,0.84 Learning_rate:0.0010 Iter:1840 Loss:0.245 Accuracy:0.96,0.92,0.84,1.00 Learning_rate:0.0010 Iter:1860 Loss:0.257 Accuracy:0.88,0.96,0.80,0.84 Learning_rate:0.0010 Iter:1880 Loss:0.331 Accuracy:1.00,0.80,0.80,0.84 Learning_rate:0.0010 Iter:1900 Loss:0.406 Accuracy:0.84,0.88,0.84,0.80 Learning_rate:0.0010 Iter:1920 Loss:0.213 Accuracy:0.92,0.92,0.96,0.84 Learning_rate:0.0010 Iter:1940 Loss:0.176 Accuracy:0.96,0.92,0.96,0.92 Learning_rate:0.0010 Iter:1960 Loss:0.237 Accuracy:0.96,0.92,0.92,0.88 Learning_rate:0.0010 Iter:1980 Loss:0.316 Accuracy:0.92,0.96,0.88,0.92 Learning_rate:0.0010 Iter:2000 Loss:0.352 Accuracy:0.88,0.92,0.84,0.88 Learning_rate:0.0003 Iter:2020 Loss:0.248 Accuracy:0.92,0.96,0.84,0.96 Learning_rate:0.0003 Iter:2040 Loss:0.172 Accuracy:0.92,0.88,0.96,0.92 Learning_rate:0.0003 Iter:2060 Loss:0.249 Accuracy:0.92,0.88,0.88,0.96 Learning_rate:0.0003 Iter:2080 Loss:0.091 Accuracy:1.00,1.00,0.92,1.00 Learning_rate:0.0003 Iter:2100 Loss:0.201 Accuracy:0.96,0.96,0.96,0.96 Learning_rate:0.0003 Iter:2120 Loss:0.119 Accuracy:1.00,0.92,0.92,0.96 Learning_rate:0.0003 Iter:2140 Loss:0.197 Accuracy:0.96,0.92,0.88,0.96 Learning_rate:0.0003 Iter:2160 Loss:0.161 Accuracy:1.00,0.92,0.96,0.88 Learning_rate:0.0003 Iter:2180 Loss:0.179 Accuracy:1.00,0.96,0.92,0.80 Learning_rate:0.0003 Iter:2200 Loss:0.063 Accuracy:0.96,1.00,1.00,1.00 Learning_rate:0.0003 Iter:2220 Loss:0.143 Accuracy:0.96,0.96,0.84,1.00 Learning_rate:0.0003 Iter:2240 Loss:0.158 Accuracy:0.96,0.92,0.96,0.92 Learning_rate:0.0003 Iter:2260 Loss:0.131 Accuracy:1.00,0.92,0.96,0.88 Learning_rate:0.0003 Iter:2280 Loss:0.144 Accuracy:0.96,1.00,0.96,0.96 Learning_rate:0.0003 Iter:2300 Loss:0.256 Accuracy:0.88,0.88,0.88,0.96 Learning_rate:0.0003 Iter:2320 Loss:0.248 Accuracy:0.96,0.88,0.88,0.92 Learning_rate:0.0003 Iter:2340 Loss:0.178 Accuracy:1.00,0.92,0.92,0.92 Learning_rate:0.0003 Iter:2360 Loss:0.191 Accuracy:0.92,0.92,1.00,0.96 Learning_rate:0.0003 Iter:2380 Loss:0.217 Accuracy:0.80,1.00,0.92,1.00 Learning_rate:0.0003 Iter:2400 Loss:0.202 Accuracy:0.88,1.00,0.92,0.84 Learning_rate:0.0003 Iter:2420 Loss:0.128 Accuracy:1.00,1.00,0.84,1.00 Learning_rate:0.0003 Iter:2440 Loss:0.190 Accuracy:0.96,0.92,1.00,0.84 Learning_rate:0.0003 Iter:2460 Loss:0.156 Accuracy:0.92,0.92,0.88,1.00 Learning_rate:0.0003 Iter:2480 Loss:0.081 Accuracy:1.00,1.00,0.92,1.00 Learning_rate:0.0003 Iter:2500 Loss:0.111 Accuracy:0.96,1.00,0.96,0.96 Learning_rate:0.0003 Iter:2520 Loss:0.092 Accuracy:1.00,0.96,0.96,1.00 Learning_rate:0.0003 Iter:2540 Loss:0.116 Accuracy:1.00,0.92,1.00,0.88 Learning_rate:0.0003 Iter:2560 Loss:0.145 Accuracy:1.00,0.80,0.96,1.00 Learning_rate:0.0003 Iter:2580 Loss:0.149 Accuracy:0.96,0.96,1.00,0.92 Learning_rate:0.0003 Iter:2600 Loss:0.157 Accuracy:1.00,0.96,0.92,0.88 Learning_rate:0.0003 Iter:2620 Loss:0.099 Accuracy:1.00,1.00,0.96,0.88 Learning_rate:0.0003 Iter:2640 Loss:0.044 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0003 Iter:2660 Loss:0.046 Accuracy:1.00,1.00,0.96,0.96 Learning_rate:0.0003 Iter:2680 Loss:0.051 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0003 Iter:2700 Loss:0.229 Accuracy:0.92,0.92,0.92,0.96 Learning_rate:0.0003 Iter:2720 Loss:0.136 Accuracy:0.96,0.96,0.96,0.92 Learning_rate:0.0003 Iter:2740 Loss:0.074 Accuracy:1.00,0.96,1.00,0.96 Learning_rate:0.0003 Iter:2760 Loss:0.301 Accuracy:0.88,0.96,0.96,0.76 Learning_rate:0.0003 Iter:2780 Loss:0.042 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0003 Iter:2800 Loss:0.081 Accuracy:1.00,0.96,0.96,0.96 Learning_rate:0.0003 Iter:2820 Loss:0.045 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0003 Iter:2840 Loss:0.061 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0003 Iter:2860 Loss:0.052 Accuracy:0.96,1.00,1.00,0.96 Learning_rate:0.0003 Iter:2880 Loss:0.122 Accuracy:0.88,1.00,1.00,0.96 Learning_rate:0.0003 Iter:2900 Loss:0.163 Accuracy:0.96,0.92,0.84,1.00 Learning_rate:0.0003 Iter:2920 Loss:0.098 Accuracy:1.00,0.92,0.96,0.96 Learning_rate:0.0003 Iter:2940 Loss:0.072 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0003 Iter:2960 Loss:0.210 Accuracy:0.88,0.96,0.96,1.00 Learning_rate:0.0003 Iter:2980 Loss:0.096 Accuracy:1.00,0.92,0.92,1.00 Learning_rate:0.0003 Iter:3000 Loss:0.043 Accuracy:1.00,0.96,0.96,1.00 Learning_rate:0.0003 Iter:3020 Loss:0.039 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0003 Iter:3040 Loss:0.149 Accuracy:0.96,0.96,1.00,0.84 Learning_rate:0.0003 Iter:3060 Loss:0.099 Accuracy:0.96,0.96,1.00,0.96 Learning_rate:0.0003 Iter:3080 Loss:0.102 Accuracy:0.96,0.96,0.96,1.00 Learning_rate:0.0003 Iter:3100 Loss:0.133 Accuracy:0.96,0.96,0.96,0.92 Learning_rate:0.0003 Iter:3120 Loss:0.032 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0003 Iter:3140 Loss:0.104 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0003 Iter:3160 Loss:0.070 Accuracy:0.92,1.00,1.00,0.96 Learning_rate:0.0003 Iter:3180 Loss:0.109 Accuracy:1.00,0.96,0.92,0.96 Learning_rate:0.0003 Iter:3200 Loss:0.128 Accuracy:1.00,1.00,0.92,0.92 Learning_rate:0.0003 Iter:3220 Loss:0.106 Accuracy:0.96,1.00,1.00,0.80 Learning_rate:0.0003 Iter:3240 Loss:0.118 Accuracy:0.96,1.00,0.96,0.96 Learning_rate:0.0003 Iter:3260 Loss:0.064 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0003 Iter:3280 Loss:0.045 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0003 Iter:3300 Loss:0.151 Accuracy:0.92,0.88,0.96,1.00 Learning_rate:0.0003 Iter:3320 Loss:0.042 Accuracy:0.96,0.96,1.00,1.00 Learning_rate:0.0003 Iter:3340 Loss:0.054 Accuracy:1.00,1.00,1.00,0.88 Learning_rate:0.0003 Iter:3360 Loss:0.143 Accuracy:1.00,0.92,0.92,0.96 Learning_rate:0.0003 Iter:3380 Loss:0.054 Accuracy:0.96,0.96,1.00,1.00 Learning_rate:0.0003 Iter:3400 Loss:0.058 Accuracy:1.00,1.00,0.92,1.00 Learning_rate:0.0003 Iter:3420 Loss:0.058 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0003 Iter:3440 Loss:0.074 Accuracy:1.00,0.92,0.96,1.00 Learning_rate:0.0003 Iter:3460 Loss:0.037 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0003 Iter:3480 Loss:0.077 Accuracy:0.96,0.92,1.00,1.00 Learning_rate:0.0003 Iter:3500 Loss:0.136 Accuracy:0.88,1.00,1.00,0.92 Learning_rate:0.0003 Iter:3520 Loss:0.034 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0003 Iter:3540 Loss:0.040 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0003 Iter:3560 Loss:0.058 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0003 Iter:3580 Loss:0.094 Accuracy:0.96,0.96,0.96,0.96 Learning_rate:0.0003 Iter:3600 Loss:0.089 Accuracy:1.00,0.96,0.92,0.96 Learning_rate:0.0003 Iter:3620 Loss:0.018 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0003 Iter:3640 Loss:0.026 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0003 Iter:3660 Loss:0.071 Accuracy:0.96,0.96,1.00,0.96 Learning_rate:0.0003 Iter:3680 Loss:0.055 Accuracy:1.00,0.92,1.00,1.00 Learning_rate:0.0003 Iter:3700 Loss:0.053 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0003 Iter:3720 Loss:0.118 Accuracy:1.00,0.96,1.00,0.96 Learning_rate:0.0003 Iter:3740 Loss:0.038 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0003 Iter:3760 Loss:0.085 Accuracy:1.00,0.92,0.96,0.96 Learning_rate:0.0003 Iter:3780 Loss:0.068 Accuracy:0.96,0.96,0.96,1.00 Learning_rate:0.0003 Iter:3800 Loss:0.129 Accuracy:0.96,0.96,1.00,0.88 Learning_rate:0.0003 Iter:3820 Loss:0.145 Accuracy:0.92,0.96,0.92,0.96 Learning_rate:0.0003 Iter:3840 Loss:0.009 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0003 Iter:3860 Loss:0.032 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0003 Iter:3880 Loss:0.088 Accuracy:0.96,0.96,0.96,0.96 Learning_rate:0.0003 Iter:3900 Loss:0.051 Accuracy:1.00,0.96,1.00,0.96 Learning_rate:0.0003 Iter:3920 Loss:0.018 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0003 Iter:3940 Loss:0.106 Accuracy:1.00,0.92,1.00,0.96 Learning_rate:0.0003 Iter:3960 Loss:0.071 Accuracy:1.00,1.00,0.92,0.96 Learning_rate:0.0003 Iter:3980 Loss:0.066 Accuracy:1.00,1.00,0.96,0.92 Learning_rate:0.0003 Iter:4000 Loss:0.038 Accuracy:1.00,0.96,1.00,0.96 Learning_rate:0.0001 Iter:4020 Loss:0.016 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4040 Loss:0.049 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:4060 Loss:0.019 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4080 Loss:0.052 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0001 Iter:4100 Loss:0.039 Accuracy:0.96,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4120 Loss:0.033 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4140 Loss:0.057 Accuracy:0.96,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4160 Loss:0.014 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4180 Loss:0.028 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4200 Loss:0.060 Accuracy:1.00,0.96,0.96,1.00 Learning_rate:0.0001 Iter:4220 Loss:0.034 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4240 Loss:0.042 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4260 Loss:0.022 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4280 Loss:0.034 Accuracy:1.00,0.92,1.00,1.00 Learning_rate:0.0001 Iter:4300 Loss:0.019 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4320 Loss:0.036 Accuracy:0.96,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4340 Loss:0.025 Accuracy:0.96,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4360 Loss:0.069 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:4380 Loss:0.018 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4400 Loss:0.038 Accuracy:1.00,0.96,0.96,1.00 Learning_rate:0.0001 Iter:4420 Loss:0.072 Accuracy:1.00,0.88,1.00,1.00 Learning_rate:0.0001 Iter:4440 Loss:0.019 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4460 Loss:0.079 Accuracy:0.96,1.00,0.92,1.00 Learning_rate:0.0001 Iter:4480 Loss:0.057 Accuracy:1.00,0.96,0.96,1.00 Learning_rate:0.0001 Iter:4500 Loss:0.023 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4520 Loss:0.027 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4540 Loss:0.010 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4560 Loss:0.032 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4580 Loss:0.011 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4600 Loss:0.052 Accuracy:0.96,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4620 Loss:0.019 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4640 Loss:0.010 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4660 Loss:0.025 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4680 Loss:0.034 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4700 Loss:0.007 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4720 Loss:0.023 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4740 Loss:0.057 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0001 Iter:4760 Loss:0.023 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4780 Loss:0.015 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4800 Loss:0.042 Accuracy:0.96,1.00,0.96,1.00 Learning_rate:0.0001 Iter:4820 Loss:0.055 Accuracy:1.00,0.92,1.00,1.00 Learning_rate:0.0001 Iter:4840 Loss:0.016 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4860 Loss:0.011 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4880 Loss:0.009 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4900 Loss:0.014 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:4920 Loss:0.056 Accuracy:0.96,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4940 Loss:0.032 Accuracy:1.00,1.00,1.00,0.92 Learning_rate:0.0001 Iter:4960 Loss:0.032 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:4980 Loss:0.027 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5000 Loss:0.003 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5020 Loss:0.018 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5040 Loss:0.059 Accuracy:0.92,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5060 Loss:0.013 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:5080 Loss:0.015 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:5100 Loss:0.009 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5120 Loss:0.012 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5140 Loss:0.053 Accuracy:1.00,1.00,0.96,0.96 Learning_rate:0.0001 Iter:5160 Loss:0.036 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5180 Loss:0.065 Accuracy:0.96,0.96,0.96,1.00 Learning_rate:0.0001 Iter:5200 Loss:0.056 Accuracy:1.00,0.96,1.00,0.92 Learning_rate:0.0001 Iter:5220 Loss:0.024 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5240 Loss:0.022 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:5260 Loss:0.009 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5280 Loss:0.031 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:5300 Loss:0.012 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5320 Loss:0.038 Accuracy:1.00,0.96,1.00,0.96 Learning_rate:0.0001 Iter:5340 Loss:0.032 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:5360 Loss:0.016 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:5380 Loss:0.012 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5400 Loss:0.011 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5420 Loss:0.010 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5440 Loss:0.009 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5460 Loss:0.003 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5480 Loss:0.028 Accuracy:0.96,1.00,0.96,1.00 Learning_rate:0.0001 Iter:5500 Loss:0.047 Accuracy:0.96,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5520 Loss:0.001 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5540 Loss:0.008 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5560 Loss:0.048 Accuracy:1.00,1.00,0.96,0.96 Learning_rate:0.0001 Iter:5580 Loss:0.027 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:5600 Loss:0.001 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5620 Loss:0.026 Accuracy:0.96,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5640 Loss:0.050 Accuracy:0.96,1.00,0.96,1.00 Learning_rate:0.0001 Iter:5660 Loss:0.030 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:5680 Loss:0.017 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5700 Loss:0.026 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:5720 Loss:0.012 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5740 Loss:0.037 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:5760 Loss:0.016 Accuracy:1.00,0.96,1.00,1.00 Learning_rate:0.0001 Iter:5780 Loss:0.002 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5800 Loss:0.020 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5820 Loss:0.021 Accuracy:1.00,1.00,1.00,0.96 Learning_rate:0.0001 Iter:5840 Loss:0.017 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5860 Loss:0.025 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:5880 Loss:0.022 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0001 Iter:5900 Loss:0.025 Accuracy:1.00,1.00,0.96,0.96 Learning_rate:0.0001 Iter:5920 Loss:0.011 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5940 Loss:0.004 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5960 Loss:0.002 Accuracy:1.00,1.00,1.00,1.00 Learning_rate:0.0001 Iter:5980 Loss:0.044 Accuracy:0.96,1.00,1.00,1.00 Learning_rate:0.0001 Iter:6000 Loss:0.020 Accuracy:1.00,1.00,0.96,1.00 Learning_rate:0.0000
4.用训练好的模型进行预测
import os import tensorflow as tf from PIL import Image from nets import nets_factory import numpy as np import matplotlib.pyplot as plt
# 不同字符数量
CHAR_SET_LEN = 10
# 图片高度
IMAGE_HEIGHT = 60
# 图片宽度
IMAGE_WIDTH = 160
# 批次
BATCH_SIZE = 1
# tfrecord文件存放路径
TFRECORD_FILE = "D:/Tensorflow/captcha/test.tfrecords"
# placeholder
x = tf.placeholder(tf.float32, [None, 224, 224])
# 从tfrecord读出数据
def read_and_decode(filename):
# 根据文件名生成一个队列
filename_queue = tf.train.string_input_producer([filename])
reader = tf.TFRecordReader()
# 返回文件名和文件
_, serialized_example = reader.read(filename_queue)
features = tf.parse_single_example(serialized_example,
features={
'image' : tf.FixedLenFeature([], tf.string),
'label0': tf.FixedLenFeature([], tf.int64),
'label1': tf.FixedLenFeature([], tf.int64),
'label2': tf.FixedLenFeature([], tf.int64),
'label3': tf.FixedLenFeature([], tf.int64),
})
# 获取图片数据
image = tf.decode_raw(features['image'], tf.uint8)
# 没有经过预处理的灰度图
image_raw = tf.reshape(image, [224, 224])
# tf.train.shuffle_batch必须确定shape
image = tf.reshape(image, [224, 224])
# 图片预处理
image = tf.cast(image, tf.float32) / 255.0
image = tf.subtract(image, 0.5)
image = tf.multiply(image, 2.0)
# 获取label
label0 = tf.cast(features['label0'], tf.int32)
label1 = tf.cast(features['label1'], tf.int32)
label2 = tf.cast(features['label2'], tf.int32)
label3 = tf.cast(features['label3'], tf.int32)
return image, image_raw, label0, label1, label2, label3# 获取图片数据和标签
image, image_raw, label0, label1, label2, label3 = read_and_decode(TFRECORD_FILE)
#使用shuffle_batch可以随机打乱
image_batch, image_raw_batch, label_batch0, label_batch1, label_batch2, label_batch3 = tf.train.shuffle_batch(
[image, image_raw, label0, label1, label2, label3], batch_size = BATCH_SIZE,
capacity = 50000, min_after_dequeue=10000, num_threads=1)
#定义网络结构
train_network_fn = nets_factory.get_network_fn(
'alexnet_v2',
num_classes=CHAR_SET_LEN,
weight_decay=0.0005,
is_training=False)
with tf.Session() as sess:
# inputs: a tensor of size [batch_size, height, width, channels]
X = tf.reshape(x, [BATCH_SIZE, 224, 224, 1])
# 数据输入网络得到输出值
logits0,logits1,logits2,logits3,end_points = train_network_fn(X)
# 预测值
predict0 = tf.reshape(logits0, [-1, CHAR_SET_LEN])
predict0 = tf.argmax(predict0, 1)
predict1 = tf.reshape(logits1, [-1, CHAR_SET_LEN])
predict1 = tf.argmax(predict1, 1)
predict2 = tf.reshape(logits2, [-1, CHAR_SET_LEN])
predict2 = tf.argmax(predict2, 1)
predict3 = tf.reshape(logits3, [-1, CHAR_SET_LEN])
predict3 = tf.argmax(predict3, 1)
# 初始化
sess.run(tf.global_variables_initializer())
# 载入训练好的模型
saver = tf.train.Saver()
saver.restore(sess,'./captcha/models/crack_captcha.model-6000')
# 创建一个协调器，管理线程
coord = tf.train.Coordinator()
# 启动QueueRunner, 此时文件名队列已经进队
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
for i in range(10):
# 获取一个批次的数据和标签
b_image, b_image_raw, b_label0, b_label1 ,b_label2 ,b_label3 = sess.run([image_batch,
image_raw_batch,
label_batch0,
label_batch1,
label_batch2,
label_batch3])
# 显示图片
img=Image.fromarray(b_image_raw[0],'L')
plt.imshow(img)
plt.axis('off')
plt.show()
# 打印标签
print('label:',b_label0, b_label1 ,b_label2 ,b_label3)
# 预测
label0,label1,label2,label3 = sess.run([predict0,predict1,predict2,predict3], feed_dict={x: b_image})
# 打印预测值
print('predict:',label0,label1,label2,label3)
# 通知其他线程关闭
coord.request_stop()
# 其他所有线程关闭之后，这一函数才能返回
coord.join(threads)
