wagons-images-classification.py

from xml.etree import ElementTree as ET

import datasets

_CITATION = """\
@InProceedings{huggingface:dataset,
title = {wagons-images-classification},
author = {TrainingDataPro},
year = {2023}
}
"""

_DESCRIPTION = """\
The dataset consists of images depicting **loaded and unloaded** wagons.
The data are organasied in two folders for loaded and unloaded wagons and assisted with
.CSV file containing text classification of the images.
This dataset can be useful for various tasks, such as *image classification, object
detection and data-driven analyses related to wagon loading and unloading processes.
The dataset is useful for **rail transport sphere**, it can be utilised for automation
the identification and classification of the wagons and further optimization of the
processes in the industry.
"""

_NAME = "wagons-images-classification"

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

_LICENSE = ""

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

_LABELS = ["loaded", "unloaded"]


class WagonsImagesClassification(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("int32"),
                    "name": datasets.Value("string"),
                    "image": datasets.Image(),
                    "label": datasets.ClassLabel(
                        num_classes=len(_LABELS),
                        names=_LABELS,
                    ),
                }
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation=_CITATION,
        )

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

    def _generate_examples(self, images):
        for idx, ((image_path, image)) in enumerate(images):
            label = "unloaded" if "unloaded" in image_path else "loaded"
            
            yield idx, {
                "id": idx,
                "name": image_path,
                "image": {"path": image_path, "bytes": image.read()},
                "label": label,
            }