coco_stuff.py

import json
import datasets
from pathlib import Path

_HOMEPAGE = 'https://cocodataset.org/'
_LICENSE = 'Creative Commons Attribution 4.0 License'
_DESCRIPTION = 'COCO is a large-scale object detection, segmentation, and captioning dataset.'
_CITATION = '''\
@article{cocodataset,
  author    = {Tsung{-}Yi Lin and Michael Maire and Serge J. Belongie and Lubomir D. Bourdev and Ross B. Girshick and James Hays and Pietro Perona and Deva Ramanan and Piotr Doll{'{a} }r and C. Lawrence Zitnick},
  title     = {Microsoft {COCO:} Common Objects in Context},
  journal   = {CoRR},
  volume    = {abs/1405.0312},
  year      = {2014},
  url       = {http://arxiv.org/abs/1405.0312},
  archivePrefix = {arXiv},
  eprint    = {1405.0312},
  timestamp = {Mon, 13 Aug 2018 16:48:13 +0200},
  biburl    = {https://dblp.org/rec/bib/journals/corr/LinMBHPRDZ14},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}
'''
_NAMES = [
    'banner', 'blanket', 'branch', 'bridge', 'building-other', 'bush',
    'cabinet', 'cage', 'cardboard', 'carpet', 'ceiling-other', 'ceiling-tile',
    'cloth', 'clothes', 'clouds', 'counter', 'cupboard', 'curtain',
    'desk-stuff', 'dirt', 'door-stuff', 'fence', 'floor-marble',
    'floor-other', 'floor-stone', 'floor-tile', 'floor-wood', 'flower', 'fog',
    'food-other', 'fruit', 'furniture-other', 'grass', 'gravel', 'ground-other',
    'hill', 'house', 'leaves', 'light', 'mat', 'metal', 'mirror-stuff',
    'moss', 'mountain', 'mud', 'napkin', 'net', 'paper', 'pavement',
    'pillow', 'plant-other', 'plastic', 'platform', 'playingfield',
    'railing', 'railroad', 'river', 'road', 'rock', 'roof', 'rug', 'salad',
    'sand', 'sea', 'shelf', 'sky-other', 'skyscraper', 'snow', 'solid-other',
    'stairs', 'stone', 'straw', 'structural-other', 'table', 'tent',
    'textile-other', 'towel', 'tree', 'vegetable', 'wall-brick', 'wall-concrete',
    'wall-other', 'wall-panel', 'wall-stone', 'wall-tile', 'wall-wood',
    'water-other', 'waterdrops', 'window-blind', 'window-other', 'wood',
    'other'
]


class COCOStuffConfig(datasets.BuilderConfig):
    '''Builder Config for coco2017'''

    def __init__(
        self, description, homepage,
        annotation_urls, **kwargs
    ):
        super(COCOStuffConfig, self).__init__(
            version=datasets.Version('1.0.0', ''),
            **kwargs
        )
        self.description = description
        self.homepage = homepage
        url = 'http://images.cocodataset.org/zips/'
        self.train_image_url = url + 'train2017.zip'
        self.val_image_url = url + 'val2017.zip'
        self.train_annotation_urls = annotation_urls['train']
        self.val_annotation_urls = annotation_urls['validation']


class COCOStuff(datasets.GeneratorBasedBuilder):
    BUILDER_CONFIGS = [
        COCOStuffConfig(
            description=_DESCRIPTION,
            homepage=_HOMEPAGE,
            annotation_urls={
                'train': 'data/stuff_train.zip',
                'validation': 'data/stuff_validation.zip'
            },
        )
    ]

    def _info(self):
        features = datasets.Features({
            'image': datasets.Image(mode='RGB', decode=True, id=None),
            'categories': datasets.Sequence(
                feature=datasets.ClassLabel(names=_NAMES),
                length=-1, id=None
            ),
            'sem.rles': datasets.Sequence(
                feature={
                    'size': datasets.Sequence(
                        feature=datasets.Value(dtype='int32', id=None),
                        length=2, id=None
                    ),
                    'counts': datasets.Value(dtype='string', id=None)
                },
                length=-1, id=None
            ),
        })
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION
        )

    def _split_generators(self, dl_manager):
        train_image_path = dl_manager.download_and_extract(
            self.config.train_image_url
        )
        val_image_path = dl_manager.download_and_extract(
            self.config.val_image_url
        )
        train_annotation_paths = dl_manager.download_and_extract(
            self.config.train_annotation_urls
        )
        val_annotation_paths = dl_manager.download_and_extract(
            self.config.val_annotation_urls
        )
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    'image_path': f'{train_image_path}/train2017',
                    'annotation_path': f'{train_annotation_paths}/stuff_train.jsonl'
                }
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    'image_path': f'{val_image_path}/val2017',
                    'annotation_path': f'{val_annotation_paths}/stuff_validation.jsonl'
                }
            )
        ]

    def _generate_examples(self, image_path, annotation_path):
        idx = 0
        image_path = Path(image_path)
        with open(annotation_path, 'r', encoding='utf-8') as f:
            for line in f:
                obj = json.loads(line.strip())
                example = {
                    'image': str(image_path / obj['image']),
                    'categories': obj['categories'],
                    'sem.rles': obj['sem.rles']
                }
                yield idx, example
                idx += 1