app.py

import logging
import time
from pathlib import Path
import contextlib

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s - %(levelname)s - %(message)s",
)


import gradio as gr
import nltk
import torch

from pdf2text import *

_here = Path(__file__).parent

nltk.download("stopwords")  # TODO=find where this requirement originates from


def load_uploaded_file(file_obj, temp_dir: Path = None):
    """
    load_uploaded_file - process an uploaded file

    Args:
        file_obj (POTENTIALLY list): Gradio file object inside a list

    Returns:
        str, the uploaded file contents
    """

    # check if mysterious file object is a list
    if isinstance(file_obj, list):
        file_obj = file_obj[0]
    file_path = Path(file_obj.name)

    if temp_dir is None:
        _temp_dir = _here / "temp"
    _temp_dir.mkdir(exist_ok=True)

    try:
        pdf_bytes_obj = open(file_path, "rb").read()
        temp_path = temp_dir / file_path.name if temp_dir else file_path
        # save to PDF file
        with open(temp_path, "wb") as f:
            f.write(pdf_bytes_obj)
        logging.info(f"Saved uploaded file to {temp_path}")
        return str(temp_path.resolve())

    except Exception as e:
        logging.error(f"Trying to load file with path {file_path}, error: {e}")
        print(f"Trying to load file with path {file_path}, error: {e}")
        return None


def convert_PDF(pdf_obj, language: str = "en", max_pages=20,):
    """
    convert_PDF - convert a PDF file to text

    Args:
        pdf_bytes_obj (bytes): PDF file contents
        language (str, optional): Language to use for OCR. Defaults to "en".

    Returns:
        str, the PDF file contents as text
    """

    global ocr_model
    st = time.perf_counter()
    if isinstance(pdf_obj, list):
        pdf_obj = pdf_obj[0]
    file_path = Path(pdf_obj.name)

    conversion_stats = convert_PDF_to_Text(
        file_path,
        ocr_model=ocr_model,
        max_pages=max_pages,
    )
    converted_txt = conversion_stats["converted_text"]
    num_pages = conversion_stats["num_pages"]
    was_truncated = conversion_stats["truncated"]
    # if alt_lang: # TODO: fix this

    rt = round((time.perf_counter() - st) / 60, 2)
    print(f"Runtime: {rt} minutes")
    html = ""
    if was_truncated:
        html += f"<p>WARNING - PDF was truncated to {max_pages} pages</p>"
    html += f"<p>Runtime: {rt} minutes on CPU for {num_pages} pages</p>"

    return converted_txt, html


if __name__ == "__main__":
    logging.info("Starting app")

    use_GPU = torch.cuda.is_available()
    logging.info(f"Using GPU status: {use_GPU}")
    logging.info("Loading OCR model")
    with contextlib.redirect_stdout(None):
        ocr_model = ocr_predictor(
            "db_resnet50",
            "crnn_mobilenet_v3_large",
            pretrained=True,
            assume_straight_pages=True,
        )

    # define pdf bytes as None
    pdf_obj = _here / "example_file.pdf"
    pdf_obj = str(pdf_obj.resolve())
    _temp_dir = _here / "temp"
    _temp_dir.mkdir(exist_ok=True)

    logging.info("starting demo")
    demo = gr.Blocks()

    with demo:

        gr.Markdown("# PDF to Text")
        gr.Markdown("A basic demo of pdf-to-text conversion using OCR from the [doctr](https://mindee.github.io/doctr/index.html) package")
        gr.Markdown("---")

        with gr.Column():

            gr.Markdown("## Load Inputs")
            gr.Markdown("Upload your own file:")
            gr.Markdown("_If no file is uploaded, a sample PDF will be used_")


            uploaded_file = gr.File(
                    label="Upload a PDF file",
                    file_count="single",
                    type="file",
                    value= _here / "example_file.pdf",
                )
                # load_file_button = gr.Button("Load Uploaded File")

            gr.Markdown("---")

        with gr.Column():
            gr.Markdown("## Convert PDF to Text")
            convert_button = gr.Button("Convert PDF!")
            out_placeholder = gr.HTML("<p><em>Output will appear below:</em></p>")
            gr.Markdown("### Output")
            OCR_text = gr.Textbox(
                label="OCR Result", placeholder="The OCR text will appear here"
            )
            text_file = gr.File(
                label="Download Text File",
                file_count="single",
                type="file",
                interactive=False,
            )
        # load_file_button.click(
        #     fn=load_uploaded_file, inputs=uploaded_file, outputs=[pdf_obj]
        # )

        convert_button.click(
            fn=convert_PDF, inputs=[uploaded_file], outputs=[OCR_text, out_placeholder]
        )
    demo.launch(enable_queue=True)