def captcha2text(height=CAPTCHA_HEIGHT, width=CAPTCHA_WIDTH):
'''
验证码图片转化为文本
:param image_list:
:param height:
:param width:
:return:
'''
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, RESTORE_PATH)
success = 0
error = 0
for num in range(0, len(os.listdir(IMAGE_TEST_PATH))):
text, image = read_dir_image_for_position(num, IMAGE_TEST_PATH)
image = convert2gray(image)
image = image.flatten() / 255
image = [image]
predict = tf.argmax(
tf.reshape(
y_conv,
[-1, CAPTCHA_LEN, len(CAPTCHA_LIST)]), 2)
vector_list = sess.run(predict, feed_dict={x: image, keep_prob: 1})
vector_list = vector_list.tolist()
pre_text = [vec2text(vector) for vector in vector_list]
if (text == ''.join(pre_text)):
success += 1
else:
error += 1
print('Label:', text, ' Predict:', pre_text)
print('total:', success + error, ',success:', success, ',error:',
error)
def getCaptcha(self):
captcha = self.sess.get(captchaUrl, headers=accessHeaders)
image = BytesIO(captcha.content)
img = Image.open(image)
imgArray = convert2gray(np.array(img)).flatten() / 255
captcha_text = self.capthca_break.predictByArray(imgArray)
captcha_text = ''.join(captcha_text)
return captcha_text
def predictaByPath(self, img):
'''通过图片路径来预测验证码'''
img = np.array(Image.open(img))
img = convert2gray(img).flatten() / 255
pred = self.pred.eval(feed_dict={self._x: [img], self._prob: 1})
captcha_text = []
for i, c in enumerate(pred):
for j in range(MAX_CAPTCHA):
captcha_text.append(chr(getIdx(c[j])))
return captcha_text
def image2text(image, height=CAPTCHA_HEIGHT, width=CAPTCHA_WIDTH):
pre_text = '1111'
tf.reset_default_graph()
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, RESTORE_PATH)
image = convert2gray(image)
image = image.flatten() / 255
image = [image]
predict = tf.argmax(
tf.reshape(y_conv,
[-1, CAPTCHA_LEN, len(CAPTCHA_LIST)]), 2)
vector_list = sess.run(predict, feed_dict={x: image, keep_prob: 1})
vector_list = vector_list.tolist()
pre_text = [vec2text(vector) for vector in vector_list]
return pre_text
def getImage(self):
count = 1
while count < 100000:
try:
captcha = self.sess.get(captchaUrl, headers=accessHeaders)
image = BytesIO(captcha.content)
img = Image.open(image)
imgArray = convert2gray(np.array(img)).flatten() / 255
captcha_text = self.capthca_break.predictByArray(imgArray)
captcha_text = ''.join(captcha_text)
self.data['captcha'] = captcha_text
res = self.sess.post(
url=loginUrl, data=self.data, headers=loginHeaders)
msg = re.findall('showmsg."(.*)",', res.text)
print(msg)
if len(msg) == 0:
path = 'data/' + captcha_text + '.png'
if not os.path.exists(path):
img.save(path)
print('img', str(count) + '.png', '预测值', captcha_text)
count += 1
except requests.exceptions.ConnectionError as e:
print(e)
if not isdir('./model'):
print('Model directory does not exists.')
return
x = placeholder(float32, [None, height * width])
keep_prob = placeholder(float32)
y_conv = cnn_graph(x, keep_prob, (height, width))
saver = Saver()
with Session() as sess:
saver.restore(sess, latest_checkpoint('./model/'))
predict = argmax(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]
return text_list
if __name__ == '__main__':
text, image = gen_captcha_text_and_image()
img = Image.fromarray(image)
image = convert2gray(image)
image = image.flatten() / 255
pre_text = captcha2text([image])
print('Text:', text, ', Actual Text:', pre_text, ', Result:',
pre_text[0] == text)
img.show()
with tf.Session() as sess:
saver.restore(sess, tf.train.latest_checkpoint('model/'))
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]
return text_list
if __name__ == '__main__':
# text, image = gen_captcha_text_and_image()
# img = Image.fromarray(image)
# image = convert2gray(image)
# image = image.flatten() / 255
# pre_text = captcha2text([image])
# print("验证码正确值:", text, ' 模型预测值:', pre_text , '正确与否:',pre_text[0] == text)
# img.show()
image = Image.open('./validImage/1.jpg')
im = np.array(image)
im = convert2gray(im)
im = im.flatten() / 255
pre_text = captcha2text([im])
print("验证码正确值:", '4R62', ' 模型预测值:', pre_text[0], '正确与否:',
pre_text[0] == '4R62')