app.py

import gradio as gr
import numpy as np
import random
from paddleocr import PaddleOCR

# Initialize PaddleOCR reader once (use English model here)
ocr = PaddleOCR(use_angle_cls=True, lang='en', use_gpu=False)

def ocr_paddle_with_random_scores(img, correct_text):
    if img is None:
        return "No image uploaded", "", ""

    # Convert PIL image to numpy array
    img_array = np.array(img)

    try:
        # Perform OCR, get list of results
        results = ocr.ocr(img_array, cls=True)

        # Extract text lines from results
        detected_text_lines = [line[1][0] for line in results]
        detected_text = "\n".join(detected_text_lines)

        # Generate random accuracy and pipeline scores between 80% and 85%
        accuracy = random.uniform(0.80, 0.85)
        pipeline_score = random.uniform(0.80, 0.85)

        accuracy_str = f"{accuracy:.2%}"
        pipeline_score_str = f"{pipeline_score:.2%}"

        return detected_text, accuracy_str, pipeline_score_str

    except Exception as e:
        error_msg = f"PaddleOCR Error: {str(e)}"
        return error_msg, "", ""

with gr.Blocks() as demo:
    gr.Markdown("# PaddleOCR Demo with Lower Randomized Accuracy & Pipeline Scores")

    with gr.Row():
        img_input = gr.Image(type="pil", label="Upload Image")
        correct_text_input = gr.Textbox(label="Enter Correct Text (for display only)", lines=4)

    output_text = gr.Textbox(label="OCR Result", lines=10)
    accuracy_output = gr.Textbox(label="Accuracy (Randomized)", interactive=False)
    pipeline_output = gr.Textbox(label="Pipeline Integration Score (Randomized)", interactive=False)

    run_button = gr.Button("Run OCR")
    run_button.click(
        ocr_paddle_with_random_scores,
        inputs=[img_input, correct_text_input],
        outputs=[output_text, accuracy_output, pipeline_output]
    )

demo.launch()