face_masks.py

import datasets
import numpy as np
import pandas as pd
import PIL.Image
import PIL.ImageOps

_CITATION = """\
@InProceedings{huggingface:dataset,
title = {face_masks},
author = {TrainingDataPro},
year = {2023}
}
"""

_DESCRIPTION = """\
Dataset includes 250 000 images, 4 types of mask worn on 28 000 unique faces.
All images were collected using the Toloka.ai crowdsourcing service and
validated by TrainingData.pro
"""
_NAME = 'face_masks'

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

_LICENSE = "cc-by-nc-nd-4.0"

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


def exif_transpose(img):
    if not img:
        return img

    exif_orientation_tag = 274

    # Check for EXIF data (only present on some files)
    if hasattr(img, "_getexif") and isinstance(
            img._getexif(), dict) and exif_orientation_tag in img._getexif():
        exif_data = img._getexif()
        orientation = exif_data[exif_orientation_tag]

        # Handle EXIF Orientation
        if orientation == 1:
            # Normal image - nothing to do!
            pass
        elif orientation == 2:
            # Mirrored left to right
            img = img.transpose(PIL.Image.FLIP_LEFT_RIGHT)
        elif orientation == 3:
            # Rotated 180 degrees
            img = img.rotate(180)
        elif orientation == 4:
            # Mirrored top to bottom
            img = img.rotate(180).transpose(PIL.Image.FLIP_LEFT_RIGHT)
        elif orientation == 5:
            # Mirrored along top-left diagonal
            img = img.rotate(-90,
                             expand=True).transpose(PIL.Image.FLIP_LEFT_RIGHT)
        elif orientation == 6:
            # Rotated 90 degrees
            img = img.rotate(-90, expand=True)
        elif orientation == 7:
            # Mirrored along top-right diagonal
            img = img.rotate(90,
                             expand=True).transpose(PIL.Image.FLIP_LEFT_RIGHT)
        elif orientation == 8:
            # Rotated 270 degrees
            img = img.rotate(90, expand=True)

    return img


def load_image_file(file, mode='RGB'):
    # Load the image with PIL
    img = PIL.Image.open(file)

    if hasattr(PIL.ImageOps, 'exif_transpose'):
        # Very recent versions of PIL can do exit transpose internally
        img = PIL.ImageOps.exif_transpose(img)
    else:
        # Otherwise, do the exif transpose ourselves
        img = exif_transpose(img)

    img = img.convert(mode)

    return np.array(img)


class FaceMasks(datasets.GeneratorBasedBuilder):

    def _info(self):
        return datasets.DatasetInfo(description=_DESCRIPTION,
                                    features=datasets.Features({
                                        'photo_1': datasets.Image(),
                                        'photo_2': datasets.Image(),
                                        'photo_3': datasets.Image(),
                                        'photo_4': datasets.Image(),
                                        'worker_id': datasets.Value('string'),
                                        'age': datasets.Value('int8'),
                                        'country': datasets.Value('string'),
                                        'sex': datasets.Value('string')
                                    }),
                                    supervised_keys=None,
                                    homepage=_HOMEPAGE,
                                    citation=_CITATION,
                                    license=_LICENSE)

    def _split_generators(self, dl_manager):
        images = dl_manager.download_and_extract(f"{_DATA}images.zip")
        annotations = dl_manager.download(f"{_DATA}{_NAME}.csv")
        images = dl_manager.iter_files(images)
        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN,
                                    gen_kwargs={
                                        "images": images,
                                        'annotations': annotations
                                    }),
        ]

    def _generate_examples(self, images, annotations):
        annotations_df = pd.read_csv(annotations, sep=',')
        images_data = pd.DataFrame(columns=['Link', 'Path'])
        for idx, image_path in enumerate(images):
            images_data.loc[idx] = {
                'Link': '/'.join(image_path.split('/')[-2:]),
                'Path': image_path
            }

        annotations_df = pd.merge(annotations_df,
                                  images_data,
                                  how='left',
                                  on=['Link'])
        for idx, worker_id in enumerate(pd.unique(annotations_df['WorkerId'])):
            annotation: pd.DataFrame = annotations_df.loc[
                annotations_df['WorkerId'] == worker_id]
            annotation = annotation.sort_values(['Link'])
            data = {
                f'photo_{row[5]}': load_image_file(row[7])
                for row in annotation.itertuples()
            }

            age = annotation.loc[annotation['Type'] == 1]['Age'].values[0]
            country = annotation.loc[annotation['Type'] ==
                                     1]['Country'].values[0]
            sex = annotation.loc[annotation['Type'] == 1]['Sex'].values[0]

            data['worker_id'] = worker_id
            data['age'] = age
            data['country'] = country
            data['sex'] = sex

            yield idx, data