app.py

# -*- coding: utf-8 -*-
"""Untitled1.ipynb
Automatically generated by Colaboratory.
Original file is located at
    https://colab.research.google.com/drive/1J4fCr7TGzdFvkCeikMAQ5af5ml2Q83W0
"""

import os
os.system('pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu')
import os, glob, fitz
import cv2
import os
import PIL
import torch
import pandas as pd
import numpy as np
import gradio as gr
from tqdm import tqdm
from scipy import ndimage
from PIL import Image, ImageDraw, ImageFont



def unnormalize_box(bbox, width, height):
     #print('shape is: ', np.asarray(bbox).shape, ' and box has values: ', bbox)
     return [
         width * (bbox[0] / 1000),
         height * (bbox[1] / 1000),
         width * (bbox[2] / 1000),
         height * (bbox[3] / 1000),
     ]

def imageconversion(pdffile):
  doc = fitz.open(pdffile)
  page = doc.load_page(0)
  zoom = 2    # zoom factor
  mat = fitz.Matrix(zoom, zoom)
  pix = page.get_pixmap(matrix = mat,dpi = 300)
  image = Image.frombytes("RGB", [pix.width, pix.height], pix.samples) 
  t=pix.save("page.jpg")
  # img = removeBorders(image)
  # noise_img = add_noise(np.array(image))
  # image = Image.fromarray(noise_img)
  return image



def completepreprocess(pdffile):
    t=imageconversion(pdffile)
    image = t.convert("RGB")
    width,height=image.size
    if ocr_type == "PaddleOCR":
        words, boxes = process_image_PaddleOCR(image, width, height)
    elif ocr_type == "Pytesseract":
        words, boxes = process_image_pytesseract(image, width, height)
  myDataFrame = pd.DataFrame()
  a=[]
  doc = fitz.open(pdffile)
  for i in range(0,len(doc)):
    page = doc.load_page(i)
    zoom = 2    # zoom factor
    mat = fitz.Matrix(zoom, zoom)
    pix = page.get_pixmap(matrix = mat,dpi = 200)
    t=pix.save("page"+str(i)+".jpg")
    images = Image.open("page"+str(i)+".jpg")
    image = images.convert("RGB")
    bbox, preds, words, image = process_image(image)
    im, df = visualize_image(bbox, preds, words, image)
    im1 = im.save("page"+str(i)+".jpg")
    a.append("page"+str(i)+".jpg")
    pred_list = []
    for number in preds:
      pred_list.append(iob_to_label(number))
    _bbox, _preds, _words = process_form(pred_list, words, bbox)
    print('page: ' + str(i) + '  ' + str(len(_preds))+ '  ' + str(len(_words)))
    df = createDataframe(_preds, _words)
    myDataFrame=myDataFrame.append(df)

  im2=mergeImageVertical(a)  
  return im2,myDataFrame


title = "OCR outputs"
description = ""

css = """.output_image, .input_image {height: 600px !important}"""
#examples = [["461BHH69.PDF"],["AP-481-RF.PDF"],["DP-095-ML.PDF"],["DQ-231-LL.PDF"],["FK-941-ET.PDF"], ["FL-078-NH.PDF"]
#              ,["14ZZ69.PDF"],["74BCA69.PDF"],["254BEG69.PDF"],["761BJQ69.PDF"],["AB-486-EH.PDF"],["AZ-211-ZA.PDF"], ["CY-073-YV.PDF"]]
# ["744BJQ69.PDF"], ['tarros_2.jpg'],

iface = gr.Interface(fn=completepreprocess,
                     #inputs=gr.inputs.Image(type="pil",optional=True,label="upload file"),
                     inputs=[
                        gr.inputs.File(label="PDF"),
                        gr.inputs.Dropdown(label="Select the Open Source OCR", choices=["PaddleOCR", "Pytesseract"]),
                    ],
                     #inputs=gr.inputs.Image(type="pil")
                     outputs=[gr.outputs.Image(type="pil", label="annotated image"),"dataframe"] ,
                     title=title,
                     description=description,
                     #examples=examples,
                     css=css,
                     analytics_enabled = True, enable_queue=True)

iface.launch(inline=False , debug=True)