pcbm_survey.py

import datasets
import json
from string import Template
from pathlib import Path

_HOMEPAGE = ""
_CITATION = ""
_LICENSE = ""
_DESCRIPTION_TEMPLATE = Template(
    "$num_classes-way image classification task "
    "to test domain shift of class $spurious_class from "
    "source context $source_context to a target context without $source_context "
    "Selected classes: $selected_classes"
)
_REPO = "https://huggingface.co/datasets/dgcnz/pcbm_survey/resolve/main"


class PCBMSurveyConfig(datasets.BuilderConfig):
    """Builder Config for PCBMSurvey"""

    def __init__(
        self,
        metadata_path: str,
        selected_classes: list[str],
        spurious_class: str,
        source_context: str,
        **kwargs,
    ):
        super(PCBMSurveyConfig, self).__init__(
            version=datasets.Version("1.0.0"), **kwargs
        )
        self.metadata_path = metadata_path
        self.selected_classes = selected_classes
        self.spurious_class = spurious_class
        self.source_context = source_context


class PCBMSurvey(datasets.GeneratorBasedBuilder):
    """PCBM Metashift Survey Images dataset"""

    """
    task_1_bed_dog.json | airplane, bed, car, cow, keyboard |  bed(dog)
    task_2_keyboard_cat.json | beach, bus, airplane, keyboard, bird | keyboard(cat)
    task_3_bed_cat.json | beach, car, airplane, bed, bird | bed(cat) 
    task_4_couch_cat.json | beach, motorcycle, airplane, couch, bird | couch(cat)
    task_5_painting_lamp.json | bus, painting, cat, computer, snowboard | painting(lamp)
    task_6_pillow_clock.json | bus, pillow, cat, computer, snowboard | pillow(clock)
    task_7_television_fireplace.json | bus, television, cat, computer, snowboard | television(fireplace)
    task_8_fork_tomato.json | car, fork, table, bed, computer | fork(tomato) 
    task_9_car_snow.json | dog, car, airplane, couch, bird | car(snow)
    """

    BUILDER_CONFIGS = [
        PCBMSurveyConfig(
            name="task_1_bed_dog",
            metadata_path="task_1_bed_dog.json",
            selected_classes=["airplane", "bed", "car", "cow", "keyboard"],
            spurious_class="bed",
            source_context="dog",
        ),
        PCBMSurveyConfig(
            name="task_2_keyboard_cat",
            metadata_path="task_2_keyboard_cat.json",
            selected_classes=["beach", "bus", "airplane", "keyboard", "bird"],
            spurious_class="keyboard",
            source_context="cat",
        ),
        PCBMSurveyConfig(
            name="task_3_bed_cat",
            metadata_path="task_3_bed_cat.json",
            selected_classes=["beach", "car", "airplane", "bed", "bird"],
            spurious_class="bed",
            source_context="cat",
        ),
        PCBMSurveyConfig(
            name="task_4_couch_cat",
            metadata_path="task_4_couch_cat.json",
            selected_classes=["beach", "motorcycle", "airplane", "couch", "bird"],
            spurious_class="couch",
            source_context="cat",
        ),
        PCBMSurveyConfig(
            name="task_5_painting_lamp",
            metadata_path="task_5_painting_lamp.json",
            selected_classes=["bus", "painting", "cat", "computer", "snowboard"],
            spurious_class="painting",
            source_context="lamp",
        ),
        PCBMSurveyConfig(
            name="task_6_pillow_clock",
            metadata_path="task_6_pillow_clock.json",
            selected_classes=["bus", "pillow", "cat", "computer", "snowboard"],
            spurious_class="pillow",
            source_context="clock",
        ),
        PCBMSurveyConfig(
            name="task_7_television_fireplace",
            metadata_path="task_7_television_fireplace.json",
            selected_classes=["bus", "television", "cat", "computer", "snowboard"],
            spurious_class="television",
            source_context="fireplace",
        ),
        PCBMSurveyConfig(
            name="task_8_fork_tomato",
            metadata_path="task_8_fork_tomato.json",
            selected_classes=["car", "fork", "table", "bed", "computer"],
            spurious_class="fork",
            source_context="tomato",
        ),
        PCBMSurveyConfig(
            name="task_9_car_snow",
            metadata_path="task_9_car_snow.json",
            selected_classes=["dog", "car", "airplane", "couch", "bird"],
            spurious_class="car",
            source_context="snow",
        ),
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION_TEMPLATE.substitute(
                num_classes=len(self.config.selected_classes),
                spurious_class=self.config.spurious_class,
                source_context=self.config.source_context,
                selected_classes=", ".join(self.config.selected_classes),
            ),
            features=datasets.Features(
                {
                    "image": datasets.Image(),
                    "label": datasets.ClassLabel(names=self.config.selected_classes),
                }
            ),
            supervised_keys=("image", "label"),
            homepage=_HOMEPAGE,
            citation=_CITATION,
            license=_LICENSE,
            task_templates=[
                datasets.ImageClassification(image_column="image", label_column="label")
            ],
        )

    def _split_generators(self, dl_manager):
        archive_path = dl_manager.download(f"{_REPO}/data/images.tar.gz")
        metadata_path = dl_manager.download(f"{_REPO}/scenarios/{self.config.metadata_path}")
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "images": dl_manager.iter_archive(archive_path),
                    "metadata_path": metadata_path,
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "images": dl_manager.iter_archive(archive_path),
                    "metadata_path": metadata_path,
                    "split": "test",
                },
            ),
        ]

    def _generate_examples(self, images, metadata_path: str, split: str):
        """Generate images and labels for splits."""
        with open(metadata_path, encoding="utf-8") as f:
            metadata = json.load(f)
            split_data = metadata["data_splits"][split]
            ids_to_keep = set()
            for _, ids in split_data.items():
                ids_to_keep.update([Path(id).stem for id in ids])

        files = dict()
        for file_path, file_obj in images:
            image_id = Path(file_path).stem
            if image_id in ids_to_keep:
                files[image_id] = (file_obj.read(), file_path)

        for cls, ids in split_data.items():
            for image_id in ids:
                image_id = Path(image_id).stem
                file_obj, file_path = files[image_id]
                yield f"{cls}_{image_id}", {
                    "image": {"path": file_path, "bytes": file_obj},
                    "label": cls,
                }