def multi_test(height=CAPTCHA_HEIGHT, width=CAPTCHA_WIDTH):
x = tf.placeholder(tf.float32, [None, height * width])
keep_prob = tf.placeholder(tf.float32)
y_conv = cnn_graph(x, keep_prob, (height, width))
saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('.'))
while 1:
text, image = get_random_captcha_text_and_image()
image = convert2gray(image)
image = image.flatten() / 255
image_list = [image]
predict = tf.argmax(
tf.reshape(
y_conv,
[-1, CAPTCHA_LEN, len(CAPTCHA_LIST)]), 2)
vector_list = sess.run(predict,
feed_dict={
x: image_list,
keep_prob: 1
})
vector_list = vector_list.tolist()
text_list = [vec2text(vector) for vector in vector_list]
pre_text = text_list[0]
flag = u'错误'
if text == pre_text:
flag = u'正确'
print u"实际值(actual):%s, 预测值(predict):%s, 预测结果:%s" % (
text,
pre_text,
flag,
)
def test_crack_captcha(test_step):
output = crack_captcha_cnn()
saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('./models'))
predict = tf.argmax(
tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
sum_correct = 0
# for _ in range(test_step):
for g_image in gen_image('./captcha_image'):
text_source, image = g_image
# print(text_source, image)
image = convert2gray(image)
captcha_image = image.flatten() / 255
text_list = sess.run(predict,
feed_dict={
X: [captcha_image],
keep_prob: 1
})
# text_list = sess.run(predict, feed_dict={X: np.reshape(captcha_image,(-1,4000)), keep_prob: 1})
text = text_list[0].tolist()
vector = np.zeros(MAX_CAPTCHA * CHAR_SET_LEN)
i = 0
for n in text:
vector[i * CHAR_SET_LEN + n] = 1
i += 1
predict_text = vec2text(vector)
print("正确: {} 预测: {}".format(text_source, predict_text))
if text_source.lower() == predict_text.lower():
sum_correct += 1
print('sum_correct:%s' % sum_correct)
print('成功率：%s' % (sum_correct / test_step))
def crack_captcha():
output = crack_captcha_cnn()
saver = tf.train.Saver()
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('models'))
predict = tf.argmax(
tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
count_out = 0
right_number = 0
while (True):
root_text, image = gen_captcha_text_and_image()
image = convert2gray(image)
captcha_image = image.flatten() / 255
text_list = sess.run(predict,
feed_dict={
X: [captcha_image],
keep_prob: 1
})
text = text_list[0].tolist()
vector = np.zeros(MAX_CAPTCHA * CHAR_SET_LEN)
i = 0
for n in text:
vector[i * CHAR_SET_LEN + n] = 1
i += 1
predict_text = vec2text(vector)
print("正确: {} 预测: {}".format(root_text, predict_text))
if count_out == 1000:
break
count_out += 1
def test_captcha_model(test_step):
sum_correct = 0
for _ in range(test_step):
# text, image = gen_captcha_text_and_image()
text, image = gen_image('./captcha_image')
image = convert2gray(image)
image = image.flatten() / 255
predict_text = crack_captcha(image)
print("正确: {} 预测: {}".format(text, predict_text))
if text == predict_text:
sum_correct += 1
print('sum_correct:%s' % sum_correct)
print('成功率：%s' % (sum_correct / test_step))
def get_name_and_image():
list1 = glob.glob(r"C:\Users\Administrator\Desktop\text\text\*.png")
num =random.randint(0,len(list1)-1)
path = list1[num]
print(path)
p, image_text = path.split('_')
captcha_text, type = image_text.split('.')
captcha_image = Image.open(path)
captcha_image = np.array(captcha_image)
image = convert2gray(captcha_image)
image = image.flatten() / 255
return captcha_text, image
def test():
img_base64 = request.form['img_base64']
# return img_base64
img_base64 = re.sub('^data:image/.+;base64,', '', img_base64)
img_base64 = base64.b64decode(img_base64)
buffer = cStringIO.StringIO(img_base64)
img = Image.open(buffer)
img = np.array(img)
img = convert2gray(img)
img = img.flatten() / 255
text_list = sess.run(pre, feed_dict={X: [img], keep_prop: 1})
text = text_list[0].tolist()
vector = np.zeros(6 * 36)
i = 0
for n in text:
vector[i * 36 + n] = 1
i += 1
text = vec2text(vector)
res = {"code": 200, "data": text, "msg": None}
return json.dumps(res)
flag = u'错误'
if text == pre_text:
flag = u'正确'
print u"实际值(actual):%s, 预测值(predict):%s, 预测结果:%s" % (
text,
pre_text,
flag,
)
if __name__ == '__main__':
try:
# 多个测试
multi_test()
exit()
text, image = get_random_captcha_text_and_image()
image = convert2gray(image)
image = image.flatten() / 255
pre_text = captcha_to_text([image])
flag = u'错误'
if text == pre_text:
flag = u'正确'
print u"实际值(actual):%s, 预测值(predict):%s, 预测结果:%s" % (
text,
pre_text,
flag,
)
except KeyboardInterrupt as e:
print e.message
from gen_captcha import gen_captcha_text_and_image
from gen_captcha import number
from gen_captcha import alphabet
from gen_captcha import ALPHABET
IMAGE_HEIGHT = 60
IMAGE_WIDTH = 160
MAX_CAPTCHA = 4
# 文本转向量
char_set = number + alphabet + ALPHABET + ['_'] # 如果验证码长度小于4, '_'用来补齐
CHAR_SET_LEN = len(char_set)
def crack_captcha(captcha_image):
output = crack_captcha_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, "./Model/model.ckpt")
text_list = sess.run(predict, feed_dict={X: [captcha_image], keep_prob: 1})
text = text_list[0].tolist()
return text
text, image = gen_captcha_text_and_image()
image = convert2gray(image)
image = image.flatten() / 255
predict_text = crack_captcha(image)
print("正确: {} 预测: {}".format(text, predict_text))
