app.py

import os
import numpy as np
import json
import shutil
import requests
import re as r
from urllib.request import urlopen
from datetime import datetime
import gradio as gr
import tensorflow as tf
import keras_ocr
import cv2
import csv
import pandas as pd
import huggingface_hub
from huggingface_hub import Repository, upload_file
import scipy.ndimage.interpolation as inter
import easyocr
from datasets import load_dataset, Image
from PIL import Image as PILImage
from paddleocr import PaddleOCR
import pytesseract
import torch
import spaces

# Global Variables
HF_TOKEN = os.environ.get("HF_TOKEN")
DATASET_NAME = "image_to_text_ocr"
DATASET_REPO_URL = "https://huggingface.co/ImranzamanML/image_to_text_ocr"
DATA_FILENAME = "ocr_data.csv"
DATA_FILE_PATH = os.path.join("ocr_data", DATA_FILENAME)
DATASET_REPO_ID = "ImranzamanML/image_to_text_ocr"
REPOSITORY_DIR = "data"
LOCAL_DIR = 'data_local'
os.makedirs(LOCAL_DIR, exist_ok=True)

"""
OCR using PaddleOCR
"""
@spaces.GPU
def paddle_ocr_processor(image):
    final_text = ''
    ocr = PaddleOCR(use_gpu=True, lang='en', use_angle_cls=True)
    result = ocr.ocr(image)
    for i in range(len(result[0])):
        text = result[0][i][1][0]
        final_text += ' ' + text
    return final_text

"""
OCR using Keras OCR
"""
@spaces.GPU
def keras_ocr_processor(image):
    output_text = ''
    pipeline = keras_ocr.pipeline.Pipeline()
    images = [keras_ocr.tools.read(image)]
    predictions = pipeline.recognize(images)
    first_prediction = predictions[0]
    for text, box in first_prediction:
        output_text += ' ' + text
    return output_text

"""
OCR using EasyOCR
"""
def convert_to_grayscale(image):
    return cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

def apply_thresholding(src):
    return cv2.threshold(src, 127, 255, cv2.THRESH_TOZERO)[1]

@spaces.GPU
def easy_ocr_processor(image):
    gray_image = convert_to_grayscale(image)
    apply_thresholding(gray_image)
    cv2.imwrite('processed_image.png', gray_image)
    reader = easyocr.Reader(['th', 'en'])
    detected_text = reader.readtext('processed_image.png', paragraph="False", detail=0)
    detected_text = ''.join(detected_text)
    return detected_text

"""
Utility Functions
"""
def save_json(data, filepath):
    with open(filepath, 'w+', encoding="utf8") as f:
        json.dump(data, f)

def get_ip_address():
    try:
        response = str(urlopen('http://checkip.dyndns.com/').read())
        return r.compile(r'Address: (\d+\.\d+\.\d+\.\d+)').search(response).group(1)
    except Exception as e:
        print("Error while getting IP address -->", e)
        return ''

def fetch_location(ip_addr):
    try:
        req_data = {"ip": ip_addr, "token": "pkml123"}
        url = "https://demos.pragnakalp.com/get-ip-location"
        headers = {'Content-Type': 'application/json'}
        response = requests.post(url, headers=headers, data=json.dumps(req_data)).json()
        return response
    except Exception as e:
        print("Error while getting location -->", e)
        return {}

def log_ocr_data(method, text_output, input_image):
    print("Logging OCR data...")
    ip_address = get_ip_address()
    location_info = fetch_location(ip_address)
    timestamp = datetime.now().strftime('%Y-%m-%d %H-%M-%S')
    save_dir = os.path.join(LOCAL_DIR, timestamp)
    os.makedirs(save_dir, exist_ok=True)
    
    image_filename = os.path.join(save_dir, 'image.png')
    try:
        PILImage.fromarray(input_image).save(image_filename)
    except Exception:
        raise Exception(f"Failed to save image as file")    

    metadata_file_path = os.path.join(save_dir, 'metadata.jsonl')
    metadata = {
        'id': timestamp,
        'method': method,
        'file_name': 'image.png',
        'generated_text': text_output,
        'ip': ip_address,
        'location': location_info
    }
    save_json(metadata, metadata_file_path)

    repo_image_path = os.path.join(REPOSITORY_DIR, os.path.join(timestamp, 'image.png'))
    _ = upload_file(
        path_or_fileobj=image_filename,
        path_in_repo=repo_image_path,
        repo_id=DATASET_REPO_ID,
        repo_type='dataset',
        token=HF_TOKEN
    )

    repo_json_path = os.path.join(REPOSITORY_DIR, os.path.join(timestamp, 'metadata.jsonl'))
    _ = upload_file(
        path_or_fileobj=metadata_file_path,
        path_in_repo=repo_json_path,
        repo_id=DATASET_REPO_ID,
        repo_type='dataset',
        token=HF_TOKEN
    )

    repo.git_pull()

    url = 'http://pragnakalpdev35.pythonanywhere.com/HF_space_image_to_text'
    payload = {
        'Method': method,
        'text_output': text_output,
        'img': input_image.tolist(),
        'ip_address': ip_address,
        'loc': location_info
    }
    response = requests.post(url, json=payload)
    print("Mail status code:", response.status_code)

    return "***** Logs saved successfully! *****"

"""
OCR Generation
"""
def generate_ocr_text(method, image):
    text_output = ''
    if image.any():
        if method == 'EasyOCR':
            text_output = easy_ocr_processor(image)
        elif method == 'KerasOCR':
            text_output = keras_ocr_processor(image)
        elif method == 'PaddleOCR':
            text_output = paddle_ocr_processor(image)

        try:
            log_ocr_data(method, text_output, image)
        except Exception as e:
            print(e)
        return text_output
    else:
        raise gr.Error("Please upload an image!")

"""
Create user interface for OCR demo
"""
image_input = gr.Image(label="Upload Image")
method_input = gr.Radio(["PaddleOCR", "EasyOCR", "KerasOCR"], value="PaddleOCR", label="Select OCR Method")
output_textbox = gr.Textbox(label="Recognized Text")

demo = gr.Interface(
    fn=generate_ocr_text,
    inputs=[method_input, image_input],
    outputs=output_textbox,
    title="Enhanced OCR Demo",
    description="Choose an OCR method and upload an image to extract text.",
    theme="huggingface",
    css="""
    .gradio-container {background-color: #f5f5f5; font-family: Arial, sans-serif;}
    #method_input {background-color: #FFC107; font-size: 18px; padding: 10px;}
    #output_textbox {font-size: 16px; color: #333;}
    """
)

demo.launch()