app.py

import gradio as gr
import cv2
import numpy as np
import pytesseract
#pytesseract.pytesseract.tesseract_cmd = "/usr/bin/tesseract"
pytesseract.pytesseract.tesseract_cmd = "tesseract"

from PIL import Image
from docx import Document
import io


# ---------- Step 1: Image preprocessing ----------
def preprocess_image(image):
    gray = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2GRAY)
    gray = cv2.bilateralFilter(gray, 11, 17, 17)
    edges = cv2.Canny(gray, 30, 200)

    contours, _ = cv2.findContours(edges, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
    contours = sorted(contours, key=cv2.contourArea, reverse=True)[:5]

    doc_contour = None
    for contour in contours:
        perimeter = cv2.arcLength(contour, True)
        approx = cv2.approxPolyDP(contour, 0.02 * perimeter, True)
        if len(approx) == 4:
            doc_contour = approx
            break

    if doc_contour is not None:
        pts = doc_contour.reshape(4, 2)
        rect = np.zeros((4, 2), dtype="float32")

        s = pts.sum(axis=1)
        rect[0] = pts[np.argmin(s)]
        rect[2] = pts[np.argmax(s)]

        diff = np.diff(pts, axis=1)
        rect[1] = pts[np.argmin(diff)]
        rect[3] = pts[np.argmax(diff)]

        (tl, tr, br, bl) = rect
        widthA = np.linalg.norm(br - bl)
        widthB = np.linalg.norm(tr - tl)
        heightA = np.linalg.norm(tr - br)
        heightB = np.linalg.norm(tl - bl)
        maxWidth = int(max(widthA, widthB))
        maxHeight = int(max(heightA, heightB))

        dst = np.array([
            [0, 0],
            [maxWidth - 1, 0],
            [maxWidth - 1, maxHeight - 1],
            [0, maxHeight - 1]
        ], dtype="float32")

        M = cv2.getPerspectiveTransform(rect, dst)
        warped = cv2.warpPerspective(np.array(image), M, (maxWidth, maxHeight))
        return Image.fromarray(warped)
    else:
        return image


# ---------- Step 2: OCR & Word export ----------
def extract_text(image):
    preprocessed = preprocess_image(image)
    text = pytesseract.image_to_string(preprocessed, lang="eng+ara")

    # Save as DOCX
    doc = Document()
    for line in text.splitlines():
        if line.strip():
            doc.add_paragraph(line)
    buffer = io.BytesIO()
    doc.save(buffer)
    buffer.seek(0)

    return text, (buffer, "document_output.docx")


# ---------- Step 3: Gradio UI ----------
def process_image(image):
    text, file_buffer = extract_text(image)
    return text, file_buffer


with gr.Blocks(title="AI Document OCR (Light Version)") as demo:
    gr.Markdown("## 📄 AI Document OCR (Light Version)\nUpload a scanned or skewed document, and the model will correct it and extract formatted text.")
    
    with gr.Row():
        input_image = gr.Image(type="pil", label="Upload Document")
    with gr.Row():
        output_text = gr.Textbox(label="Extracted Text", lines=15)
        output_file = gr.File(label="Download as .docx")
    with gr.Row():
        submit_btn = gr.Button("Extract Text")

    submit_btn.click(process_image, inputs=input_image, outputs=[output_text, output_file])

demo.launch()