yalta_ai_tabular_dataset.py

# Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Script for reading 'You Actually Look Twice At it (YALTAi)' dataset."""


import os
from glob import glob

import datasets
from PIL import Image

_CITATION = """\
    @dataset{clerice_thibault_2022_6827706,
  author       = {Clérice, Thibault},
  title        = {YALTAi: Tabular Dataset},
  month        = jul,
  year         = 2022,
  publisher    = {Zenodo},
  version      = {1.0.0},
  doi          = {10.5281/zenodo.6827706},
  url          = {https://doi.org/10.5281/zenodo.6827706}
}
"""

_DESCRIPTION = """Yalt AI Tabular Dataset"""

_HOMEPAGE = "https://doi.org/10.5281/zenodo.6827706"

_LICENSE = "Creative Commons Attribution 4.0 International"

_URL = "https://zenodo.org/record/6827706/files/yaltai-table.zip?download=1"

_CATEGORIES = ["Header", "Col", "Marginal", "text"]


class YaltAiTabularDatasetConfig(datasets.BuilderConfig):
    """BuilderConfig for YaltAiTabularDataset."""

    def __init__(self, name, **kwargs):
        """BuilderConfig for YaltAiTabularDataset."""
        super(YaltAiTabularDatasetConfig, self).__init__(
            version=datasets.Version("1.0.0"), name=name, description=None, **kwargs
        )


class YaltAiTabularDataset(datasets.GeneratorBasedBuilder):
    """Object Detection for historic manuscripts"""

    BUILDER_CONFIGS = [
        YaltAiTabularDatasetConfig("YOLO"),
        YaltAiTabularDatasetConfig("COCO"),
    ]

    def _info(self):
        if self.config.name == "COCO":
            features = datasets.Features(
                {
                    "image_id": datasets.Value("int64"),
                    "image": datasets.Image(),
                    "width": datasets.Value("int32"),
                    "height": datasets.Value("int32"),
                }
            )
            object_dict = {
                "category_id": datasets.ClassLabel(names=_CATEGORIES),
                "image_id": datasets.Value("string"),
                "id": datasets.Value("int64"),
                "area": datasets.Value("int64"),
                "bbox": datasets.Sequence(datasets.Value("float32"), length=4),
                "segmentation": [[datasets.Value("float32")]],
                "iscrowd": datasets.Value("bool"),
            }
            features["objects"] = [object_dict]
        if self.config.name == "YOLO":
            features = datasets.Features(
                {
                    "image": datasets.Image(),
                    "objects": datasets.Sequence(
                        {
                            "label": datasets.ClassLabel(names=_CATEGORIES),
                            "bbox": datasets.Sequence(
                                datasets.Value("int32"), length=4
                            ),
                        }
                    ),
                }
            )
        return datasets.DatasetInfo(
            features=features,
            supervised_keys=None,
            description=_DESCRIPTION,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        data_dir = dl_manager.download_and_extract(_URL)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_dir": os.path.join(data_dir, "yaltai-table/", "train")
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={"data_dir": os.path.join(data_dir, "yaltai-table/", "val")},
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "data_dir": os.path.join(data_dir, "yaltai-table/", "test")
                },
            ),
        ]

    def _generate_examples(self, data_dir):
        def create_annotation_from_yolo_format(
            min_x,
            min_y,
            width,
            height,
            image_id,
            category_id,
            annotation_id,
            segmentation=False,
        ):
            bbox = (float(min_x), float(min_y), float(width), float(height))
            area = width * height
            max_x = min_x + width
            max_y = min_y + height
            if segmentation:
                seg = [[min_x, min_y, max_x, min_y, max_x, max_y, min_x, max_y]]
            else:
                seg = []
            return {
                "id": annotation_id,
                "image_id": image_id,
                "bbox": bbox,
                "area": area,
                "iscrowd": 0,
                "category_id": category_id,
                "segmentation": seg,
            }

        image_dir = os.path.join(data_dir, "images")
        label_dir = os.path.join(data_dir, "labels")
        image_paths = sorted(glob(f"{image_dir}/*.jpg"))
        label_paths = sorted(glob(f"{label_dir}/*.txt"))
        if self.config.name == "COCO":
            for idx, (image_path, label_path) in enumerate(
                zip(image_paths, label_paths)
            ):
                image_id = idx
                annotations = []
                image = Image.open(image_path)  # Possibly conver to RGB?
                w, h = image.size
                with open(label_path, "r") as f:
                    lines = f.readlines()
                for line in lines:
                    line = line.strip().split()
                    category_id = line[0]
                    x_center = float(line[1])
                    y_center = float(line[2])
                    width = float(line[3])
                    height = float(line[4])

                    float_x_center = w * x_center
                    float_y_center = h * y_center
                    float_width = w * width
                    float_height = h * height

                    min_x = int(float_x_center - float_width / 2)
                    min_y = int(float_y_center - float_height / 2)
                    width = int(float_width)
                    height = int(float_height)

                    annotation = create_annotation_from_yolo_format(
                        min_x,
                        min_y,
                        width,
                        height,
                        image_id,
                        category_id,
                        image_id,
                    )
                    annotations.append(annotation)

                example = {
                    "image_id": image_id,
                    "image": image,
                    "width": w,
                    "height": h,
                    "objects": annotations,
                }
                yield idx, example
        if self.config.name == "YOLO":
            for idx, (image_path, label_path) in enumerate(
                zip(image_paths, label_paths)
            ):
                im = Image.open(image_path)
                width, height = im.size
                image_id = idx
                annotations = []
                with open(label_path, "r") as f:
                    lines = f.readlines()
                objects = []
                for line in lines:
                    line = line.strip().split()
                    bbox_class = int(line[0])
                    bbox_xcenter = int(float(line[1]) * width)
                    bbox_ycenter = int(float(line[2]) * height)
                    bbox_width = int(float(line[3]) * width)
                    bbox_height = int(float(line[4]) * height)
                    objects.append(
                        {
                            "label": bbox_class,
                            "bbox": [
                                bbox_xcenter,
                                bbox_ycenter,
                                bbox_width,
                                bbox_height,
                            ],
                        }
                    )

                yield idx, {
                    "image": image_path,
                    "objects": objects,
                }