hl.py

# coding=utf-8
# 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.
"""High-Level dataset."""


import json
from pathlib import Path

import datasets


_CITATION = """\
@inproceedings{Cafagna2023HLDG,
  title={HL Dataset: Grounding High-Level Linguistic Concepts in Vision},
  author={Michele Cafagna and Kees van Deemter and Albert Gatt},
  year={2023}
}
"""

_DESCRIPTION = """\
High-level Dataset
"""

# github link
_HOMEPAGE = "https://github.com/michelecafagna26/HL-dataset"

_LICENSE = "Apache 2.0"

_IMG = "https://huggingface.co/datasets/michelecafagna26/hl/resolve/main/data/images.tar.gz"
_TRAIN = "https://huggingface.co/datasets/michelecafagna26/hl/resolve/main/data/annotations/train.jsonl"
_TEST = "https://huggingface.co/datasets/michelecafagna26/hl/resolve/main/data/annotations/test.jsonl"



class HL(datasets.GeneratorBasedBuilder):
    """High Level Dataset."""

    VERSION = datasets.Version("1.0.0")

    def _info(self):
        features = datasets.Features(
            {
                "file_name": datasets.Value("string"),
                "image": datasets.Image(),
                "scene": datasets.Sequence(datasets.Value("string")),
                "action": datasets.Sequence(datasets.Value("string")),
                "rationale": datasets.Sequence(datasets.Value("string")),
                "object": datasets.Sequence(datasets.Value("string")),
                "confidence": {
                        "scene": datasets.Sequence(datasets.Value("float32")),
                        "action": datasets.Sequence(datasets.Value("float32")),
                        "rationale": datasets.Sequence(datasets.Value("float32")),
                    },
                "purity": {
                    "scene": datasets.Sequence(datasets.Value("float32")),
                    "action": datasets.Sequence(datasets.Value("float32")),
                    "rationale": datasets.Sequence(datasets.Value("float32")),
                },
                "diversity": {
                    "scene": datasets.Value("float32"),
                    "action": datasets.Value("float32"),
                    "rationale": datasets.Value("float32"),
                },
            }
        )
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        image_files = dl_manager.download(_IMG)
        annotation_files = dl_manager.download_and_extract([_TRAIN, _TEST])
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "annotation_file_path": annotation_files[0],
                    "images": dl_manager.iter_archive(image_files),
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "annotation_file_path": annotation_files[1],
                    "images": dl_manager.iter_archive(image_files),
                },
            ),
        ]

    def _generate_examples(self, annotation_file_path, images):
        
        idx = 0

        #assert Path(annotation_file_path).suffix == ".jsonl"
        
        with open(annotation_file_path, "r") as fp:
            metadata = {json.loads(item)['file_name']: json.loads(item) for item in fp}

        # This loop relies on the ordering of the files in the archive:
        # Annotation files come first, then the images.
        for img_file_path, img_obj in images:

            file_name = Path(img_file_path).name
            
            if file_name in metadata:
                yield idx, {
                        "file_name": file_name,
                        "image": {"path": img_file_path, "bytes": img_obj.read()},
                        "scene": metadata[file_name]['captions']['scene'],
                        "action": metadata[file_name]['captions']['action'],
                        "rationale": metadata[file_name]['captions']['rationale'],
                        "object": metadata[file_name]['captions']['object'],
                        "confidence": metadata[file_name]['confidence'],
                        "purity": metadata[file_name]['purity'],
                        "diversity": metadata[file_name]['diversity'],
                    }
                idx += 1