suim.py

# Copyright 2020 The HuggingFace Datasets Authors and the current dataset
# script contributor.
#
# 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.
"""Semantic Segmentation of Underwater IMagery (SUIM) dataset"""


import os

import datasets


_CITATION = """\
@inproceedings{islam2020suim,
  title={{Semantic Segmentation of Underwater Imagery: Dataset and Benchmark}},
  author={Islam, Md Jahidul and Edge, Chelsey and Xiao, Yuyang and Luo, Peigen and Mehtaz, 
              Muntaqim and Morse, Christopher and Enan, Sadman Sakib and Sattar, Junaed},
  booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  year={2020},
  organization={IEEE/RSJ}
}
"""

_DESCRIPTION = """\
The SUIM dataset is a dataset for semantic segmentation of underwater imagery.

The dataset consists of 1525 annotated images for training/validation and
110 samples for testing.

| Object category                  | Symbol | RGB color code |
|----------------------------------|--------|----------------|
| Background (waterbody)           | BW     | 000 (black)    |
| Human divers                     | HD     | 001 (blue)     |
| Aquatic plants and sea-grass     | PF     | 010 (green)    |
| Wrecks and ruins                 | WR     | 011 (sky)      |
| Robots (AUVs/ROVs/instruments)   | RO     | 100 (red)      |
| Reefs and invertebrates          | RI     | 101 (pink)     |
| Fish and vertebrates             | FV     | 110 (yellow)   |
| Sea-floor and rocks              | SR     | 111 (white)    |


For more information about the original SUIM dataset,
please visit the official dataset page:

https://irvlab.cs.umn.edu/resources/suim-dataset

Please refer to the original dataset source for any additional details,
citations, or specific usage guidelines provided by the dataset creators.
"""

_HOMEPAGE = "https://irvlab.cs.umn.edu/resources/suim-dataset"

_LICENSE = "mit"


class ExDark(datasets.GeneratorBasedBuilder):
    """Semantic Segmentation of Underwater IMagery (SUIM) dataset"""

    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="suim",
            version=VERSION,
            description="Semantic Segmentation of Underwater IMagery (SUIM) dataset",
        ),
    ]

    DEFAULT_CONFIG_NAME = "suim"

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "img": datasets.Image(),
                    "mask": datasets.Image(),
                }
            ),
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        data_dir = dl_manager.download_and_extract("SUIM.zip")
        train_dir = os.path.join(data_dir, "SUIM", "train_val")
        test_dir = os.path.join(data_dir, "SUIM", "TEST")

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={
                    "data_dir": train_dir,
                    "split": "train",
                },
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={
                    "data_dir": test_dir,
                    "split": "test",
                },
            ),
        ]

    def _generate_examples(self, data_dir, split):
        img_dir = os.path.join(data_dir, "images")
        masks_dir = os.path.join(data_dir, "masks")
        img_files = os.listdir(img_dir)

        for idx, img_file in enumerate(img_files):
            img_path = os.path.join(img_dir, img_file)
            mask_path = os.path.join(
                masks_dir,
                img_file.replace(".jpg", ".bmp"),
            )
            record = {
                "img": img_path,
                "mask": mask_path,
            }
            yield idx, record