学习了这么久终于可以开始动手实现验证码破解了
不得不说卷积神经网络在图像识别方向真的是核武器级别的神器，验证码识别率高达99.87%！远超人类95%的识别率，niubility！！！
import tensorflow as tf
import numpy as np
from PIL import Image
import random, datetime
# 字符长度,验证码长度, 图片高宽
CHAR_SET_LEN = 10
MAX_CAPTCHA, IMAGE_HEIGHT, IMAGE_WIDTH = 4, 80, 200
# 日志和模型保存目录
TRAIN_LOG_PATH, TRAIN_MODEL_PATH = "logs/", "model/fuck_captche.model-1800"
# 训练库位置
TRAIN_LABLE_PATH, TRAIN_IMGS_PATH = "D:\\verifies\\train\\verfiycodes.txt", "D:\\verifies\\train\\"
# 测试库位置
TEST_LABLE_PATH, TEST_IMGS_PATH = "D:\\verifies\\test\\verfiycodes.txt", "D:\\verifies\\test\\"
# x y占位符
X = tf.placeholder(tf.float32, [None, IMAGE_HEIGHT * IMAGE_WIDTH])
Y = tf.placeholder(tf.float32, [None, CHAR_SET_LEN * MAX_CAPTCHA])
keep_prob = tf.placeholder(tf.float32)
def _convert2gray(img):
"""
彩色图片转灰色图片
:param img:
:return:
"""
if len(img.shape)>2:
r, g, b = img[:, :, 0], img[:, :, 1], img[:, :, 2]
gray = 0.2989 * r + 0.5870 * g + 0.1140 * b
return gray
else:
return img
def _char2pos(c):
"""
字符串pos
:param c:
:return:
"""
if c == '_':
k = 62
return k
k = ord(c) - 48
if k > 9:
k = ord(c) - 55
if k > 35:
k = ord(c) - 61
if k > 61:
raise ValueError('No Map')
return k
def _text2vec(text):
"""
文本转向量
:param text:
:return:
"""
vector = np.zeros(MAX_CAPTCHA * CHAR_SET_LEN)
for i, c in enumerate(text):
idx = i * CHAR_SET_LEN + _char2pos(c)
vector[idx] = 1
return vector
def _get_captcha_texts(path=TRAIN_LABLE_PATH):
"""
获取所有训练标签
:param path:
:return:
"""
texts = []
for i in open(path, 'r'):
texts.append(i.strip('\n'))
return texts
def _get_captcha_text_and_image(index):
"""
获取验证码和对应的标签
:param index:
:return:
"""
if IS_TRAIN:
imgs_path = TRAIN_IMGS_PATH + str(index)+".jpg"
lable = TRAIN_MODEL_TEXTS[index-1]
else:
imgs_path = TEST_IMGS_PATH + str(index)+".jpg"
lable = TEST_MODEL_TEXTS[index-1]
image = np.array(Image.open(imgs_path))
return lable, image
def _get_next_batch(batch_size=60):
"""
生成一个batch
:param batch_size:
:return:
"""
batch_x = np.zeros([BATCH, 80 * 200])
batch_y = np.zeros([BATCH, 40])
batch_index = 0
end_index = batch_size + 1
start_index = end_index - BATCH
# 测试
if IS_TRAIN:
path = TRAIN_IMGS_PATH
else:
path = TEST_IMGS_PATH
for i in range(start_index, end_index):
text, image = _get_captcha_text_and_image(i)
image = _convert2gray(image)
# 将图片数组一维化 同时将文本也对应在两个二维组的同一行
batch_x[batch_index, :] = image.flatten() / 255 # (image.flatten()-128)/128 mean为0
batch_y[batch_index, :] = _text2vec(text)
batch_index = batch_index + 1
# 返回该训练批次
return batch_x, batch_y
def cnn(b_alpha=0.1):
"""
3层卷积神经网络
:param b_alpha:
:return:
"""
x = tf.reshape(X, shape=[-1, IMAGE_HEIGHT, IMAGE_WIDTH, 1])
wc1 = tf.get_variable(name='wc1', shape=[3, 3, 1, 32],
dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer())
bc1 = tf.Variable(b_alpha * tf.random_normal([32]))
conv1 = tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(x, wc1, strides=[1, 1, 1, 1], padding='SAME'), bc1)) # 输出大小不变
conv1 = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv1 = tf.nn.dropout(conv1, keep_prob)
wc2 = tf.get_variable(name='wc2', shape=[3, 3, 32, 64],
dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer())
bc2 = tf.Variable(b_alpha * tf.random_normal([64]))
conv2 = tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(conv1, wc2, strides=[1, 1, 1, 1], padding='SAME'), bc2))
conv2 = tf.nn.max_pool(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv2 = tf.nn.dropout(conv2, keep_prob)
wc3 = tf.get_variable(name='wc3', shape=[3, 3, 64, 128],
dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer())
bc3 = tf.Variable(b_alpha * tf.random_normal([128]))
conv3 = tf.nn.relu(tf.nn.bias_add(tf.nn.conv2d(conv2, wc3, strides=[1, 1, 1, 1], padding='SAME'), bc3))
conv3 = tf.nn.max_pool(conv3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
conv3 = tf.nn.dropout(conv3, keep_prob)
# 经过三次卷积后，得到10*25大小的图片，128是上层输入的大小
wd1 = tf.get_variable(name='wd1', shape=[10*25*128, 1024],
dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer())
bd1 = tf.Variable(b_alpha * tf.random_normal([1024]))
dense = tf.reshape(conv3, [-1, 10*25*128])
dense = tf.nn.relu(tf.add(tf.matmul(dense, wd1), bd1))
dense = tf.nn.dropout(dense, keep_prob)
wout = tf.get_variable('name', shape=[1024, MAX_CAPTCHA * CHAR_SET_LEN],
dtype=tf.float32, initializer=tf.contrib.layers.xavier_initializer())
bout = tf.Variable(b_alpha * tf.random_normal([MAX_CAPTCHA * CHAR_SET_LEN]))
out = tf.add(tf.matmul(dense, wout), bout)
return out
def train():
"""
训练模型函数
:return:
"""
output = cnn()
cost = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=output, labels=Y))
optimizer = tf.train.AdamOptimizer(learning_rate=0.001).minimize(cost)
predict = tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN])
max_idx_p = tf.argmax(predict, 2)
max_idx_l = tf.argmax(tf.reshape(Y, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
correct_pred = tf.equal(max_idx_p, max_idx_l)
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
saver = tf.train.Saver()
init = tf.global_variables_initializer()
with tf.Session() as sess:
sess.run(init)
sess.graph.finalize()
for step in range(1, 2001):
start_time = datetime.datetime.now()
batch_x, batch_y = _get_next_batch(BATCH * step)
_, cost_ = sess.run([optimizer, cost],
feed_dict={X: batch_x, Y: batch_y, keep_prob: 0.75})
end_time = datetime.datetime.now()
print("step=%s, cost=%s, spending times=%.2fs"
% (step, cost_, (end_time-start_time).microseconds / 1000000))
# 每100步测试一下准确率
if step % 100 == 0:
# 测试数据集使用下一次的数据集
batch_x_test, batch_y_test = _get_next_batch(BATCH * (step+1))
acc = sess.run(accuracy,
feed_dict={X: batch_x_test, Y: batch_y_test, keep_prob: 0.75})
print("step=%d, 准确率 ------------> %s" % (step, acc))
# 达到99%准确率就保存model并退出
if acc > 0.99:
saver.save(sess, TRAIN_MODEL_PATH, global_step=step)
break
def _fuck_captcha(sess, predict, captcha_image):
"""
破解验证码方法
:param captcha_image:
:return:
"""
text_list = sess.run(predict, feed_dict={X: [captcha_image], keep_prob: 1})
text = text_list[0].tolist()
text = "".join(list(map(str, text)))
return text
def test():
"""
测试函数
:return:
"""
output = cnn()
saver = tf.train.Saver()
predict = tf.argmax(tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint("model/"))
task_count = 5000
right_count = 0
for i in range(task_count):
text, image = _get_captcha_text_and_image(i+1)
image = _convert2gray(image)
image = image.flatten() / 255
predict_text = _fuck_captcha(sess, predict, image)
if str(text) == predict_text:
right_count += 1
else:
print("【错误】: \t正确值: {} 预测值: {}".format(text, predict_text))
print('正确/共计-----', right_count, '/', task_count)
if __name__ == '__main__':
# 获取所有指定库的标签
IS_TRAIN = True
IS_TRAIN = False
TRAIN_MODEL_TEXTS, TEST_MODEL_TEXTS = [], []
if IS_TRAIN:
BATCH = 100 # 每次取batch条数据作为训练集
TRAIN_MODEL_TEXTS = _get_captcha_texts(TRAIN_LABLE_PATH)
train()
else:
BATCH = 2000
TEST_MODEL_TEXTS = _get_captcha_texts(TEST_LABLE_PATH)
test()
测试模型：5000张图片的准确率：
【错误】: 正确值: 1574 预测值: 1514
【错误】: 正确值: 1639 预测值: 1635
【错误】: 正确值: 6779 预测值: 6719
【错误】: 正确值: 5711 预测值: 5771
【错误】: 正确值: 7136 预测值: 7138
【错误】: 正确值: 7888 预测值: 7883
【错误】: 正确值: 1425 预测值: 1426
【错误】: 正确值: 8116 预测值: 3116
【错误】: 正确值: 6836 预测值: 6236
【错误】: 正确值: 2678 预测值: 2673
【错误】: 正确值: 2648 预测值: 2643
【错误】: 正确值: 9784 预测值: 9781
【错误】: 正确值: 3244 预测值: 3247
正确/共计----- 4987 / 5000
网友评论
我用20w的张图片训练的模型，能达到9997/10000的准确度。
准确率高得太吓人了。