from paddleocr import PaddleOCR, draw_ocr
from PIL import Image
import torch

OCR = PaddleOCR(use_angle_cls=True, lang='en') # need to run only once to download and load model into memory