This loader extracts the text from an image that has text in it (e.g. a receipt). The Donut transformer model is used. The file extensions .png, .jpg, and .jpeg are preferred. A single local file is passed in each time you call load_data.

Usage

To use this loader, you need to pass in a Path to a local file.

from pathlib import Path
from gpt_index import download_loader

ImageReader = download_loader("ImageReader")

loader = ImageReader()
documents = loader.load_data(file=Path('./receipt.png'))

This loader is designed to be used as a way to load data into GPT Index and/or subsequently used as a Tool in a LangChain Agent. See here for examples.