ocr-receipts-text-detection.py

from xml.etree import ElementTree as ET

import datasets

_CITATION = """\
@InProceedings{huggingface:dataset,
title = {ocr-receipts-text-detection},
author = {TrainingDataPro},
year = {2023}
}
"""

_DESCRIPTION = """\
The Grocery Store Receipts Dataset is a collection of photos captured from various
**grocery store receipts**. This dataset is specifically designed for tasks related to
**Optical Character Recognition (OCR)** and is useful for retail.
Each image in the dataset is accompanied by bounding box annotations, indicating the
precise locations of specific text segments on the receipts. The text segments are
categorized into four classes: **item, store, date_time and total**.
"""
_NAME = "ocr-receipts-text-detection"

_HOMEPAGE = f"https://huggingface.co/datasets/TrainingDataPro/{_NAME}"

_LICENSE = ""

_DATA = f"https://huggingface.co/datasets/TrainingDataPro/{_NAME}/resolve/main/data/"

_LABELS = ["receipt", "shop", "item", "date_time", "total"]


class OcrReceiptsTextDetection(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("int32"),
                    "name": datasets.Value("string"),
                    "image": datasets.Image(),
                    "mask": datasets.Image(),
                    "width": datasets.Value("uint16"),
                    "height": datasets.Value("uint16"),
                    "shapes": datasets.Sequence(
                        {
                            "label": datasets.ClassLabel(
                                num_classes=len(_LABELS),
                                names=_LABELS,
                            ),
                            "type": datasets.Value("string"),
                            "points": datasets.Sequence(
                                datasets.Sequence(
                                    datasets.Value("float"),
                                ),
                            ),
                            "rotation": datasets.Value("float"),
                            "occluded": datasets.Value("uint8"),
                            "attributes": datasets.Sequence(
                                {
                                    "name": datasets.Value("string"),
                                    "text": datasets.Value("string"),
                                }
                            ),
                        }
                    ),
                }
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        images = dl_manager.download(f"{_DATA}images.tar.gz")
        masks = dl_manager.download(f"{_DATA}boxes.tar.gz")
        annotations = dl_manager.download(f"{_DATA}annotations.xml")
        images = dl_manager.iter_archive(images)
        masks = dl_manager.iter_archive(masks)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "images": images,
                    "masks": masks,
                    "annotations": annotations,
                },
            ),
        ]

    @staticmethod
    def parse_shape(shape: ET.Element) -> dict:
        label = shape.get("label")
        shape_type = shape.tag
        rotation = shape.get("rotation", 0.0)
        occluded = shape.get("occluded", 0)

        points = None

        if shape_type == "points":
            points = tuple(map(float, shape.get("points").split(",")))

        elif shape_type == "box":
            points = [
                (float(shape.get("xtl")), float(shape.get("ytl"))),
                (float(shape.get("xbr")), float(shape.get("ybr"))),
            ]

        elif shape_type == "polygon":
            points = [
                tuple(map(float, point.split(",")))
                for point in shape.get("points").split(";")
            ]

        attributes = []

        for attr in shape:
            attr_name = attr.get("name")
            attr_text = attr.text
            attributes.append({"name": attr_name, "text": attr_text})

        shape_data = {
            "label": label,
            "type": shape_type,
            "points": points,
            "rotation": rotation,
            "occluded": occluded,
            "attributes": attributes,
        }

        return shape_data

    def _generate_examples(self, images, masks, annotations):
        tree = ET.parse(annotations)
        root = tree.getroot()

        for idx, (
            (image_path, image),
            (mask_path, mask),
        ) in enumerate(zip(images, masks)):
            image_name = image_path.split("/")[-1]
            img = root.find(f"./image[@name='images/{image_name}']")

            image_id = img.get("id")
            name = img.get("name")
            width = img.get("width")
            height = img.get("height")
            shapes = [self.parse_shape(shape) for shape in img]

            yield idx, {
                "id": image_id,
                "name": name,
                "image": {"path": image_path, "bytes": image.read()},
                "mask": {"path": mask_path, "bytes": mask.read()},
                "width": width,
                "height": height,
                "shapes": shapes,
            }