llama-hub/loader_hub/file/image/base.py

"""Image Reader.

A parser for image files.

"""

import re
from pathlib import Path
from typing import Dict, Optional, cast, List

from llama_index.readers.base import BaseReader
from llama_index.readers.schema.base import Document, ImageDocument


class ImageReader(BaseReader):
    """Image parser.

    Extract text from images using DONUT.

    """

    def __init__(
        self,
        text_type: str = "text",
        parser_config: Optional[Dict] = None,
        keep_image: bool = False,
        parse_text: bool = True,
    ):
        """Init parser."""
        self._text_type = text_type
        if parser_config is None and parse_text:
            if text_type == "plain_text":
                import pytesseract

                processor = None
                model = pytesseract
            else:
                from transformers import DonutProcessor, VisionEncoderDecoderModel

                processor = DonutProcessor.from_pretrained(
                    "naver-clova-ix/donut-base-finetuned-cord-v2"
                )
                model = VisionEncoderDecoderModel.from_pretrained(
                    "naver-clova-ix/donut-base-finetuned-cord-v2"
                )
            parser_config = {"processor": processor, "model": model}
        self._parser_config = parser_config
        self._keep_image = keep_image
        self._parse_text = parse_text

    def load_data(
        self, file: Path, extra_info: Optional[Dict] = None
    ) -> List[Document]:
        """Parse file."""
        from PIL import Image

        from llama_index.img_utils import img_2_b64

        # load document image
        image = Image.open(file)
        if image.mode != "RGB":
            image = image.convert("RGB")

        # Encode image into base64 string and keep in document
        image_str: Optional[str] = None
        if self._keep_image:
            image_str = img_2_b64(image)

        # Parse image into text
        text_str: str = ""
        if self._parse_text:
            model = self._parser_config["model"]
            processor = self._parser_config["processor"]

            if processor:
                import torch

                device = "cuda" if torch.cuda.is_available() else "cpu"
                model.to(device)

                # prepare decoder inputs
                task_prompt = "<s_cord-v2>"
                decoder_input_ids = processor.tokenizer(
                    task_prompt, add_special_tokens=False, return_tensors="pt"
                ).input_ids

                pixel_values = processor(image, return_tensors="pt").pixel_values

                outputs = model.generate(
                    pixel_values.to(device),
                    decoder_input_ids=decoder_input_ids.to(device),
                    max_length=model.decoder.config.max_position_embeddings,
                    early_stopping=True,
                    pad_token_id=processor.tokenizer.pad_token_id,
                    eos_token_id=processor.tokenizer.eos_token_id,
                    use_cache=True,
                    num_beams=3,
                    bad_words_ids=[[processor.tokenizer.unk_token_id]],
                    return_dict_in_generate=True,
                )

                sequence = processor.batch_decode(outputs.sequences)[0]
                sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(
                    processor.tokenizer.pad_token, ""
                )
                # remove first task start token
                text_str = re.sub(r"<.*?>", "", sequence, count=1).strip()
            else:
                import pytesseract

                model = cast(pytesseract, self._parser_config["model"])
                text_str = model.image_to_string(image)

        return [
            ImageDocument(
                text=text_str,
                image=image_str,
            )
        ]