app.py

import gradio as gr
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from pdf2image import convert_from_path
import pytesseract

# Load TrOCR Model from Hugging Face
processor = TrOCRProcessor.from_pretrained("microsoft/trocr-base-handwritten")
model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-handwritten")

# Function to extract text from PDF
def extract_text_from_pdf(pdf_path):
    images = convert_from_path(pdf_path)
    extracted_text = []
    
    for img in images:
        # Convert image to text using TrOCR
        pixel_values = processor(images=img, return_tensors="pt").pixel_values
        generated_ids = model.generate(pixel_values)
        text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
        
        # Fallback to Tesseract if TrOCR fails
        if not text.strip():
            text = pytesseract.image_to_string(img)
        
        extracted_text.append(text)
    
    return "\n".join(extracted_text)

# Gradio Interface
def ocr_pipeline(pdf_file):
    pdf_path = pdf_file.name
    extracted_text = extract_text_from_pdf(pdf_path)
    return extracted_text

iface = gr.Interface(
    fn=ocr_pipeline,
    inputs=gr.File(label="Upload PDF"),
    outputs="text",
    title="PDF Text Extraction using TrOCR"
)

# Run the Gradio App
if __name__ == "__main__":
    iface.launch()