def recog(imgstr, ext='.png'):
if not isinstance(imgstr, str) and len(imgstr) > 0:
return {'ok': False, 'error': 'Invalid Image: not string or too small'}
try:
img = base64.b64decode(imgstr)
except:
return {'ok': False, 'error': 'Invalid Image: not valid encode'}
path = '/tmp/'
path += str(uuid.uuid4()) + ext
with open(path, 'wb') as fp:
fp.write(img)
try:
r = tr.run(path)
r = [{
'x': int(a[0]),
'y': int(a[1]),
'w': int(a[2]),
'h': int(a[3]),
'text': b,
'c': c
} for a, b, c in r]
try:
os.remove(path)
except:
pass
return {'ok': True, 'data': r}
except:
try:
os.remove(path)
except:
pass
traceback.print_exc()
return {'ok': False, 'error': 'Parse Error'}
def test():
os.chdir(_BASEDIR)
print("recognize", tr.recognize("imgs/line.png"))
img_path = "imgs/id_card.jpeg"
# img_path = "imgs/name_card.jpg"
img_pil = Image.open(img_path)
print(img_pil.size)
color_pil = img_pil.convert("RGB")
gray_pil = img_pil.convert("L")
img_draw = ImageDraw.Draw(color_pil)
colors = ['red', 'green', 'blue', "purple"]
t = time.time()
n = 1
for _ in range(n):
tr.detect(gray_pil, flag=tr.FLAG_RECT)
print("time", (time.time() - t) / n)
results = tr.run(gray_pil, flag=tr.FLAG_ROTATED_RECT)
for i, rect in enumerate(results):
cx, cy, w, h, a = tuple(rect[0])
print(i, "\t", rect[1], rect[2])
box = cv2.boxPoints(((cx, cy), (w, h), a))
box = np.int0(np.round(box))
for p1, p2 in [(0, 1), (1, 2), (2, 3), (3, 0)]:
img_draw.line(xy=(box[p1][0], box[p1][1], box[p2][0], box[p2][1]),
fill=colors[i % len(colors)],
width=2)
color_pil.show()
def test():
print("recognize", tr.recognize("imgs/line.png"))
img_path = "imgs/id_card.jpeg"
# img_path = "imgs/name_card.jpg"
img_pil = Image.open(img_path)
try:
if hasattr(img_pil, '_getexif'):
# from PIL import ExifTags
# for orientation in ExifTags.TAGS.keys():
# if ExifTags.TAGS[orientation] == 'Orientation':
# break
orientation = 274
exif = dict(img_pil._getexif().items())
if exif[orientation] == 3:
img_pil = img_pil.rotate(180, expand=True)
elif exif[orientation] == 6:
img_pil = img_pil.rotate(270, expand=True)
elif exif[orientation] == 8:
img_pil = img_pil.rotate(90, expand=True)
except:
pass
MAX_SIZE = 1600
if img_pil.height > MAX_SIZE or img_pil.width > MAX_SIZE:
scale = max(img_pil.height / MAX_SIZE, img_pil.width / MAX_SIZE)
new_width = int(img_pil.width / scale + 0.5)
new_height = int(img_pil.height / scale + 0.5)
img_pil = img_pil.resize((new_width, new_height), Image.ANTIALIAS)
color_pil = img_pil.convert("RGB")
gray_pil = img_pil.convert("L")
img_draw = ImageDraw.Draw(color_pil)
colors = ['red', 'green', 'blue', "purple"]
t = time.time()
n = 1
for _ in range(n):
tr.detect(gray_pil, flag=tr.FLAG_RECT)
print("time", (time.time() - t) / n)
results = tr.run(gray_pil, flag=tr.FLAG_ROTATED_RECT)
for i, rect in enumerate(results):
cx, cy, w, h, a = tuple(rect[0])
print(i, "\t", rect[1], rect[2])
box = cv2.boxPoints(((cx, cy), (w, h), a))
box = np.int0(np.round(box))
for p1, p2 in [(0, 1), (1, 2), (2, 3), (3, 0)]:
img_draw.line(xy=(box[p1][0], box[p1][1], box[p2][0], box[p2][1]),
fill=colors[i % len(colors)],
width=2)
color_pil.show()
def job_ocr(self, image_bytes):
result = ''
try:
path = "./test.jpg"
pyBytes = bytes(image_bytes.data())
with open(path, 'wb') as f:
f.write(pyBytes)
result_list = tr.run(path)
if (len(result_list) != 0):
for result_single in result_list:
result += result_single[1]
finally:
self.signal_response.emit('')
self.signal_response.emit(result)
pyperclip.copy(result)
def job_ocr2(self, image_bytes):
result = ''
try:
path = "./test.jpg"
# pyBytes=bytes(image_bytes.data())
pyBytes = image_bytes
with open(path, 'wb') as f:
f.write(pyBytes)
result_list = tr.run(path)
result_list_final = get_title_from_reg_text(result_list)
if (len(result_list) != 0):
for result_single in result_list_final:
result += result_single + "\n"
finally:
self.signal_response.emit('')
self.signal_response.emit(result)
import tr
# detect text lines, return list of (x, y, w, h)
print(tr.detect("imgs/web.png"))
# recognize text line, return (text, confidence)
print(tr.recognize("imgs/line.png"))
# detect and recognize, return list of ((x, y, w, h), text, confidence)
print(tr.run("imgs/name_card.jpg"))
def post(self):
'''
:return:
报错：
400 没有请求参数
'''
start_time = time.time()
MAX_SIZE = 1600
img_up = self.request.files.get('file', None)
img_b64 = self.get_argument('img', None)
compress_size = self.get_argument('compress', None)
# 判断是上传的图片还是base64
self.set_header('content-type', 'application/json')
up_image_type = None
if img_up is not None and len(img_up) > 0:
img_up = img_up[0]
up_image_type = img_up.content_type
up_image_name = img_up.filename
img = Image.open(BytesIO(img_up.body))
elif img_b64 is not None:
raw_image = base64.b64decode(img_b64.encode('utf8'))
img = Image.open(BytesIO(raw_image))
else:
self.set_status(400)
logger.error(json.dumps({'code': 400, 'msg': '没有传入参数'}, cls=NpEncoder))
self.finish(json.dumps({'code': 400, 'msg': '没有传入参数'}, cls=NpEncoder))
return
try:
if hasattr(img, '_getexif') and img._getexif() is not None:
orientation = 274
exif = dict(img._getexif().items())
if exif[orientation] == 3:
img = img.rotate(180, expand=True)
elif exif[orientation] == 6:
img = img.rotate(270, expand=True)
elif exif[orientation] == 8:
img = img.rotate(90, expand=True)
except Exception as ex:
error_log = json.dumps({'code': 400, 'msg': '产生了一点错误，请检查日志', 'err': str(ex)}, cls=NpEncoder)
logger.error(error_log, exc_info=True)
self.finish(error_log)
return
img = img.convert("RGB")
'''
是否开启图片压缩
默认为1600px
值为 0 时表示不开启压缩
非 0 时则压缩到该值的大小
'''
if compress_size is not None:
try:
compress_size = int(compress_size)
except ValueError as ex:
logger.error(exc_info=True)
self.finish(json.dumps({'code': 400, 'msg': 'compress参数类型有误，只能是int类型'}, cls=NpEncoder))
return
if compress_size < 1:
MAX_SIZE = max(img.height, img.width)
else:
MAX_SIZE = compress_size
if img.height > MAX_SIZE or img.width > MAX_SIZE:
scale = max(img.height / MAX_SIZE, img.width / MAX_SIZE)
new_width = int(img.width / scale + 0.5)
new_height = int(img.height / scale + 0.5)
img = img.resize((new_width, new_height), Image.BICUBIC)
res = tr.run(img)
img_detected = img.copy()
img_draw = ImageDraw.Draw(img_detected)
colors = ['red', 'green', 'blue', "purple"]
for i, r in enumerate(res):
rect, txt, confidence = r
x, y, w, h = rect
for xy in [(x, y, x + w, y), (x + w, y, x + w, y + h), (x + w, y + h, x, y + h), (x, y + h, x, y)]:
img_draw.line(xy=xy, fill=colors[i % len(colors)], width=2)
output_buffer = BytesIO()
img_detected.save(output_buffer, format='JPEG')
byte_data = output_buffer.getvalue()
img_detected_b64 = base64.b64encode(byte_data).decode('utf8')
log_info = {
'ip': self.request.host,
'return': res,
'time': datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')
}
logger.info(json.dumps(log_info, cls=NpEncoder))
self.finish(json.dumps(
{'code': 200, 'msg': '成功',
'data': {'img_detected': 'data:image/jpeg;base64,' + img_detected_b64, 'raw_out': res,
'speed_time': round(time.time() - start_time, 2)}},
cls=NpEncoder))
return
def run_task(i):
x = tr.run(imgs[i % len(imgs)])
return i, len(x)
new_height = int(img_pil.height / scale + 0.5)
img_pil = img_pil.resize((new_width, new_height), Image.ANTIALIAS)
color_pil = img_pil.convert("RGB")
gray_pil = img_pil.convert("L")
img_draw = ImageDraw.Draw(color_pil)
colors = ['red', 'green', 'blue', "purple"]
t = time.time()
n = 1
for _ in range(n):
tr.detect(gray_pil, flag=tr.FLAG_RECT)
print("time", (time.time() - t) / n)
results = tr.run(gray_pil, flag=tr.FLAG_ROTATED_RECT)
for i, rect in enumerate(results):
cx, cy, w, h, a = tuple(rect[0])
print(i, "\t", rect[1], rect[2])
box = cv2.boxPoints(((cx, cy), (w, h), a))
box = np.int0(np.round(box))
for p1, p2 in [(0, 1), (1, 2), (2, 3), (3, 0)]:
img_draw.line(xy=(box[p1][0], box[p1][1], box[p2][0], box[p2][1]), fill=colors[i % len(colors)], width=2)
color_pil.show()
# In[ ]:
def test():
os.chdir(_BASEDIR)
print("recognize", tr.recognize("imgs/line.png"))
txt = tr.run("imgs/line.png")[0][1]
print(txt)
def tr_run(path, return_dict):
try:
return_dict['ret'] = tr.run(path)
except:
return_dict['ret'] = None