renovation.py

import os
import glob
import random

import datasets
from datasets.tasks import ImageClassification
from datasets import load_dataset
import os
from huggingface_hub import login
_HOMEPAGE = "https://github.com/your-github/renovation"

_CITATION = """\
@ONLINE {renovationdata,
    author="Your Name",
    title="Renovation dataset",
    month="January",
    year="2023",
    url="https://github.com/your-github/renovation"
}
"""

_DESCRIPTION = """\
Renovations is a dataset of images of houses taken in the field using smartphone
cameras. It consists of 7 classes: Not Applicable, Poor, Fair, Good, and Great  renovations.
Data was collected by the your research lab.
"""

_URLS = {
    "Not Applicable": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/Not Applicable.zip",
    "Poor": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/Poor.zip",
    "Fair": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/Fair.zip",
    "Good": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/Good.zip",
    "Great": "https://huggingface.co/datasets/rshrott/renovation/resolve/main/Great.zip",
}

_NAMES = ["Not Applicable", "Poor", "Fair", "Good", "Great"]
class Renovations(datasets.GeneratorBasedBuilder):
    """Renovations house images dataset."""

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "image_file_path": datasets.Value("string"),
                    "image": datasets.Image(),
                    "labels": datasets.features.ClassLabel(names=_NAMES),
                }
            ),
            supervised_keys=("image", "labels"),
            homepage=_HOMEPAGE,
            citation=_CITATION,
            task_templates=[ImageClassification(image_column="image", label_column="labels")],
        )

    def _split_generators(self, dl_manager):
        data_files = dl_manager.download_and_extract(_URLS)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_files": data_files,
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={
                    "data_files": data_files,
                    "split": "val",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "data_files": data_files,
                    "split": "test",
                },
            ),
        ]

    def _generate_examples(self, data_files, split):
        # Separate data by class
        data_by_class = {label: [] for label in _NAMES}
        allowed_extensions = {'.jpeg', '.jpg'}
        for label, path in data_files.items():
            files = [os.path.join(path, f) for f in os.listdir(path) if os.path.isfile(os.path.join(path, f)) and os.path.splitext(f)[1] in allowed_extensions]
            data_by_class[label].extend((file, label) for file in files)
    
    
        # Shuffle and split data for each class
        random.seed(43)  # ensure reproducibility
        train_data, test_data, val_data = [], [], []
        for label, files_and_labels in data_by_class.items():
            random.shuffle(files_and_labels)
            num_files = len(files_and_labels)
            train_end = int(num_files * 0.8)
            test_end = int(num_files * 0.9)
            train_data.extend(files_and_labels[:train_end])
            test_data.extend(files_and_labels[train_end:test_end])
            val_data.extend(files_and_labels[test_end:])
    
        # Select the appropriate split
        if split == "train":
            data_to_use = train_data
        elif split == "test":
            data_to_use = test_data
        else:  # "val" split
            data_to_use = val_data
    
        # Yield examples
        for idx, (file, label) in enumerate(data_to_use):
            yield idx, {
                "image_file_path": file,
                "image": file,
                "labels": label,
            }