def get_cell_digits(x, y, table_struct, img):
cell = cell_table_image(x, y, table_struct, img)
# cv2_imshow(cell)
# return ocr(cell, config='--oem 1 --psm 6 -c textord_heavy_nr=true', lang='digits1')
text = ocr(cell, config='--oem 1 --psm 7', lang='digits')
# text = text2num(text)
return text
def _get_captcha(session):
url = 'http://eoz.one/user/captcha?refresh=1'
r = utils.get(session, url)
url = r.json()['url']
url = 'http://eoz.one' + url
r = utils.get(session, url)
return utils.ocr(r.content)
def add_reminders(img_path):
img = open_image(img_path)
ocr_text = ocr(img)
menu_items = text_to_menu_items(ocr_text)[:4]
for date, food in menu_items:
add_reminder(date, food)
return menu_items
def get_cell_text(x, y, table_struct, img):
cell = cell_table_image(x, y, table_struct, img)
# cv2_imshow(cell)
# return ocr(cell, config='--oem 1 --psm 6 -c textord_heavy_nr=true', lang='rus2+digits1')
# return ocr(cell, config='--oem 1 --psm 6 -c textord_heavy_nr=true', lang='rus+digits1')
text = ocr(cell,
config='--oem 1 --psm 6 -c textord_heavy_nr=true',
lang='rus')
text = error_correct(text)
return text
def test_baidu_index(self):
baidu_index_page = page.BaiduIndexPage(self.driver, self.action)
baidu_index_page.openPage("https://index.baidu.com/")
baidu_index_page.input_search_key("start")
baidu_index_page.click_submit_button()
baidu_index_page.maxWindows()
'''
理想化的函数：
@param1: 地区:全国, 北京，香港
@param2: 查询日期:2018-09
@param3: 查询关键字: 赵丽颖+zhaoliyig
@output: void 处理过程中将结果转换成csv
'''
df = pd.DataFrame(
columns=['key_words', 'date_range', 'index_type', 'baidu_index'])
i = 0
for keywords in utils.setAllKeywords():
for date in utils.setAllDate():
# try:
baidu_index_page.input_new_search_key(keywords)
baidu_index_page.click_new_submit_button()
time.sleep(1)
# 开始自定义时间
self.inputSelfDefineDate(baidu_index_page, Constant.YEAR_2011,
Constant.OCT, Constant.YEAR_2012,
Constant.OCT)
baidu_index_page.click_index_average()
for indexType in ["Total", "PC", "Mobile"]:
filename = keywords + '.' + indexType + '.' + date
if "PC" == indexType:
baidu_index_page.click_pc_index_button()
if "Mobile" == indexType:
baidu_index_page.click_mobile_index_button()
baidu_index_page.hoverOnAvgIndex(3)
baidu_index_page.hoverOnAvgIndex(4)
baidu_index_page.saveThePicture(filename)
time.sleep(1)
avg_index = utils.ocr(Constant.IDENTIIFIED_PICTURE_FOLDER +
filename + ".png").replace(",", "")
df.loc[i] = [keywords, date, indexType, avg_index]
i += 1
# except Exception as e:
# print(keywords+date+"get failed")
# print(e)
# time.sleep(60*60)
df.to_csv(Constant.FINAL_RESULT_DIR + Constant.FINAL_RESULT_FILE_NAME,
index=False,
sep=',')
def classify_image(img_path):
img = open_image(img_path)
ocr_text = ocr(img)
if 'mosaicscience' in ocr_text:
return 'mosaic'
days_mentioned = [
token for line in ocr_text.split('\n') for token in line.split(' ')
if token in DAYS
]
if len(days_mentioned) > 5:
return 'food'
return 'trustnet'
def on_status(self, status, firstcall=True):
# Save ID of first tweet with @mention to reply to
if firstcall:
self.id = status.id
# One or multiple captions?
opts = ['tell me more', 'what else', 'is that all', 'can you do better']
if any(substr in status.text.lower() for substr in opts):
self.more = True
else:
self.more = False
# If tweet has an image, tag this one
if hasattr(status, 'extended_entities'):
print(status.text)
# Setup DenseCap
headers = {
'api-key': self.key,
}
# Iterate through each image in tweet
caption = []
for each in status.extended_entities['media']:
# Get image URL
img = each['media_url_https']
files = {
'image': img,
}
# Call DenseCap API
response = requests.post('https://api.deepai.org/api/densecap', headers=headers, files=files).json()
# Call tesseract for OCR
text = ocr(img)
# Construct caption
if 'output' in response:
if not self.more:
# Take first caption provided
cap = response['output']['captions'][0]['caption']
else:
# Take next 3 captions
cap = '; '.join([response['output']['captions'][i]['caption'] for i in range(1, 4)])
# Add OCR text if detected
if text is not None:
cap = 'The image shows ' + cap + ' and it says ' + text
caption.append(cap)
if caption:
# Build tweet
if len(caption) == 1:
tweet = caption[0]
else:
tweet = ''
for i, c in enumerate(caption):
tweet += f'[{i+1}] {c}\n'
print(tweet)
# Tweet (reply) the response
if len(tweet) <= 280:
self.api.update_status(tweet, in_reply_to_status_id=self.id, auto_populate_reply_metadata=True)
else:
tweets = tweetsplitter(tweet)
prev = self.id
for t in tweets:
latest = self.api.update_status(t, in_reply_to_status_id=prev, auto_populate_reply_metadata=True)
prev = latest.id
# If tweet has no image, try and find parent tweet with image
else:
if status.in_reply_to_status_id is not None:
parent = self.api.get_status(status.in_reply_to_status_id)
# Recursive call
self.on_status(parent, False)
def read_title(img):
ocr_text = ocr(img)
ocr_title = ocr_text.split('\n\n')[0]
return ocr_title
score, pos = _match(img, tp_startup)
if score > 0.9:
utils.tap((pos[0] + pos[1]) / 2, (pos[2] + pos[3]) / 2)
state.chg_state('run')
time.sleep(1)
else:
time.sleep(0.3)
elif state.state == 'run':
score, pos = _match(img, tp_main)
if score > 0.20:
simg = img[content_offset:, :, :]
print('Call ocr...', end=' ')
st = time.time()
ocr_rst = utils.ocr(simg, regions)
et = time.time()
print('Time: %.2f' % (et - st))
print('Question:', ocr_rst)
if '' not in ocr_rst:
question, opt1, opt2, opt3, opt4 = ocr_rst
with env.begin() as txn:
ans = txn.get(question.encode())
if ans is None:
select = random.randrange(1, 5)
else:
ans = ans.decode()
print('Memory hit!', ans)