def crack_captcha(captcha_image):
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)
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
return vec_to_text(vector)
predict = tf.argmax(tf.reshape(output, [-1, MAX_CAPTCHA, CHAR_SET_LEN]), 2)
for i in range(5000):
idx = str('%04d' % i)
img_name = dir_name + idx + '.jpg'
new = denoise(img_name)
NaiveRemoveNoise(new)
new.save('clean.jpg')
real, image = get_captcha_text_and_image(i, 'clean.jpg')
image = convert_to_gray(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 = vec_to_text(vector)
print('Real: {} Predict: {}'.format(real, predict_text))
file.write(idx + ',' + predict_text.upper() + '\n')
file.close()