app.py

import gradio as gr
import tensorflow as tf
import keras_ocr
import requests
import cv2
import os
import csv
import numpy as np
import pandas as pd
import huggingface_hub
from huggingface_hub import Repository
from datetime import datetime
import scipy.ndimage.interpolation as inter
import easyocr
import datasets
from datasets import load_dataset, Image
from PIL import Image
from paddleocr import PaddleOCR
import pytesseract
from pdf2image import convert_from_path
from PIL import Image
import os  
import tempfile


# Function to perform OCR
def ocr(input_file, lang='fas'):  # 'fas': Persian language (Farsi)
    extracted_text = ""
    # if isinstance(input_file, Image.Image):  # If the input is an image
    text = pytesseract.image_to_string(input_file, lang=lang)
    extracted_text = text

    return extracted_text

"""
Paddle OCR
"""
def ocr_with_paddle(img):
    finaltext = ''
    ocr = PaddleOCR(lang='fa', use_angle_cls=True)
    # img_path = 'exp.jpeg'
    result = ocr.ocr(img)
    
    for i in range(len(result[0])):
        text = result[0][i][1][0]
        finaltext += ' '+ text
    return finaltext

"""
Keras OCR
"""
def ocr_with_keras(img):
    output_text = ''
    pipeline=keras_ocr.pipeline.Pipeline()
    images=[keras_ocr.tools.read(img)]
    predictions=pipeline.recognize(images)
    first=predictions[0]
    for text,box in first:
        output_text += ' '+ text
    return output_text

"""
easy OCR
"""
# gray scale image
def get_grayscale(image):
    return cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

# Thresholding or Binarization
def thresholding(src):
    return cv2.threshold(src,127,255, cv2.THRESH_TOZERO)[1]
def ocr_with_easy(img):
    gray_scale_image=get_grayscale(img)
    thresholding(gray_scale_image)
    cv2.imwrite('image.png',gray_scale_image)
    reader = easyocr.Reader(['fa','en'])
    bounds = reader.readtext('image.png',paragraph="False",detail = 0)
    bounds = ''.join(bounds)
    return bounds

"""
Generate OCR
"""

def process(input_type, gr_img, lang):

    if isinstance(gr_img, np.ndarray):
        # Create a temporary file to save the image
        with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as temp_file:
            temp_path = temp_file.name
            # Save the NumPy array as an image file
            Image.fromarray(gr_img).save(temp_path)
    
    image = Image.open(temp_path)
    # image = file
    extracted_text = ocr(image, lang)
    return extracted_text


def generate_ocr(Method, img):
    
    text_output = ''

    add_csv = []
    image_id = 1
    print("Method___________________",Method)
    if Method == 'Version2':
        text_output = ocr_with_easy(img)
    if Method == 'KerasOCR':
        text_output = ocr_with_keras(img)
    if Method == 'Version1':
        text_output = ocr_with_paddle(img)
    if Method == 'Version3':
        text_output = process("img", img, "fas")
    return text_output


"""
Create user interface for OCR demo
"""

# image = gr.Image(shape=(300, 300))
# image = gr.File(label="Upload PDF/Image")
image = gr.Image()
method = gr.Radio(["Version1","Version2", "Version3"],value="PaddleOCR")
output = gr.Textbox(label="Output")

demo = gr.Interface(
    generate_ocr,
    [method, image],
    output,
    title="Persian Optical Character Recognition",
    css=".gradio-container {background-color: lightgray} #radio_div {background-color: #FFD8B4; font-size: 40px;}",
    article = """<p style='text-align: center;'>Feel free to give us your thoughts on this demo and please contact us at 
                    <a href="mailto:[email protected]" target="_blank">[email protected]</a> 
                    <p style='text-align: center;'>Developed by: <a href="https://github.com/mohammad2928" target="_blank">Mohammad Mahmoudi</a></p>"""
    

)
# demo.launch(enable_queue = False)
demo.launch(share=True)