app.py

import gradio as gr
import time
import numpy as np
from PIL import Image
from paddleocr import PaddleOCR
from transformers import TrOCRProcessor, VisionEncoderDecoderModel

# Initialize models
paddle_ocr = PaddleOCR(lang='fa', use_textline_orientation=True)
trocr_processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-printed")
trocr_model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-printed")

def run_paddleocr(image):
    """Run PaddleOCR on image"""
    # Convert to numpy array if needed
    if isinstance(image, Image.Image):
        image = np.array(image)
    
    result = paddle_ocr.ocr(image, cls=True)
    text = ' '.join([line[1][0] for line in result[0]]) if result else ''
    return text

def run_trocr(image):
    """Run TrOCR on image"""
    # Convert to PIL Image if needed
    if isinstance(image, np.ndarray):
        image = Image.fromarray(image)
    
    pixel_values = trocr_processor(image, return_tensors="pt").pixel_values
    generated_ids = trocr_model.generate(pixel_values)
    return trocr_processor.batch_decode(generated_ids, skip_special_tokens=True)[0]

def compare_models(image):
    """Compare PaddleOCR and TrOCR models"""
    # Convert to RGB if needed
    if isinstance(image, np.ndarray):
        image = Image.fromarray(image)
    image = image.convert("RGB")
    
    results = {}
    times = {}
    
    # Run PaddleOCR
    start = time.time()
    results['PaddleOCR'] = run_paddleocr(image)
    times['PaddleOCR'] = time.time() - start
    
    # Run TrOCR
    start = time.time()
    results['TrOCR'] = run_trocr(image)
    times['TrOCR'] = time.time() - start
    
    # Create comparison table
    comparison = f"""
    <table style="width:100%">
        <tr>
            <th style="text-align:center">مدل</th>
            <th style="text-align:center">متن استخراج شده</th>
            <th style="text-align:center">زمان پردازش (ثانیه)</th>
        </tr>
        <tr>
            <td style="text-align:center">PaddleOCR</td>
            <td style="text-align:right; direction:rtl">{results['PaddleOCR']}</td>
            <td style="text-align:center">{times['PaddleOCR']:.3f}</td>
        </tr>
        <tr>
            <td style="text-align:center">TrOCR</td>
            <td style="text-align:right; direction:rtl">{results['TrOCR']}</td>
            <td style="text-align:center">{times['TrOCR']:.3f}</td>
        </tr>
    </table>
    """
    
    return comparison, results['PaddleOCR'], results['TrOCR']

# Create Gradio interface
with gr.Blocks(title="مقایسه مدل‌های OCR فارسی") as demo:
    gr.Markdown("""
    ## مقایسه عملکرد مدل‌های OCR برای زبان فارسی
    این برنامه دو مدل مختلف OCR را روی تصاویر فارسی مقایسه می‌کند:
    1. PaddleOCR
    2. TrOCR (مایکروسافت)
    """)
    
    with gr.Row():
        with gr.Column():
            image_input = gr.Image(label="تصویر ورودی", type="pil")
            submit_btn = gr.Button("مقایسه مدل‌ها", variant="primary")
        
        with gr.Column():
            comparison_output = gr.HTML(label="نتایج مقایسه")
            paddle_output = gr.Textbox(label="PaddleOCR")
            trocr_output = gr.Textbox(label="TrOCR")
    
    submit_btn.click(
        fn=compare_models,
        inputs=image_input,
        outputs=[comparison_output, paddle_output, trocr_output]
    )

if __name__ == "__main__":
    demo.launch()