add dataset def

.gitattributes

@@ -57,3 +57,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+# Comic files - compressed
+*.cbz filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,10 @@
+# Python-generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+
+# Virtual environments
+.venv

README.md

@@ -0,0 +1,95 @@
+---
+license: mit
+task_categories:
+- image-classification
+- image-to-image
+- image-to-text
+language:
+- ja
+- en
+tags:
+- manga
+- comics
+- adult
+pretty_name: Nhentai Dataset
+size_categories:
+- 10K<n<100K
+---
+
+# Nhentai Dataset
+
+A collection of Japanese manga in CBZ format from Nhentai, containing adult
+content manga with associated metadata.
+
+## Content Warning
+
+This dataset contains adult content (NSFW) and is intended for research purposes
+only. Users must be of legal age and comply with all applicable laws and
+regulations.
+
+## Key Applications
+
+- Image Analysis: Suitable for manga page segmentation, panel detection, and art
+  style analysis
+- Text Recognition: Useful for Japanese text detection and translation in manga
+  contexts
+- Content Classification: Applicable for genre classification and content
+  tagging systems
+
+## Dataset Structure
+
+The dataset is organized as a collection of CBZ files, each representing a
+complete manga volume with associated metadata.
+
+### Data Instances
+
+Each instance represents a manga volume containing:
+
+- Sequential manga pages as images
+- Associated metadata (title, tags, etc.)
+- Text content (if available)
+
+### Data Fields
+
+- `comic`: List of images from the CBZ file
+- `title`: Full title of the work
+- `media_id`: Unique identifier
+- `num_favorites`: Number of user favorites
+- `tag`: List of content tags and themes
+- `language`: List of available languages
+- `artist`: List of credited artists
+- `category`: Content category (e.g., "manga")
+- `num_pages`: Total number of pages
+- `scanlator`: Translation group (if any)
+- `group`: List of publishing groups
+- `parody`: List of source materials
+- `character`: List of featured characters
+- `epos`: Epoch timestamp of upload
+- `file_name`: Original archive filename
+
+## Dataset Creation
+
+### Source Data
+
+The data is sourced from publicly available manga on Nhentai, processed and
+stored in CBZ format for efficient access and preservation of image quality.
+
+### Preprocessing
+
+- Images are preserved in their original resolution
+- Metadata is extracted and standardized
+- Files are organized in CBZ format for efficient storage and access
+
+## Considerations
+
+### Social Impact and Biases
+
+This dataset contains adult content and may reflect various biases. Users should
+be aware of and account for these aspects in their applications.
+
+### Legal & Ethical Considerations
+
+- Users must comply with age restrictions and local regulations
+- The dataset should be used responsibly and in accordance with applicable laws
+- Proper content filtering and access controls should be implemented in
+  applications

nhentai.py

@@ -0,0 +1,122 @@
+import io
+import zipfile
+from collections.abc import Sequence
+from dataclasses import dataclass
+from typing import TYPE_CHECKING, Union
+
+import datasets
+
+if TYPE_CHECKING:
+    import numpy as np
+    import PIL.Image
+
+logger = datasets.utils.logging.get_logger(__name__)
+
+
+@dataclass
+class Comic(datasets.Image):
+    """Comic feature that extends Image feature for CBZ files."""
+
+    def encode_example(
+        self, value: Sequence[Union[str, bytes, dict, "np.ndarray", "PIL.Image.Image"]]
+    ) -> dict:
+        """Encode example into a format for Arrow.
+
+        Args:
+            value (`Sequence` of `str`, `bytes`, `dict`, `np.ndarray`, or
+            `PIL.Image.Image`):
+                Sequence of data passed as input to Comic feature. Each element can be:
+                - A path to a local image file
+                - Raw bytes of an image file
+                - A dictionary containing image data
+                - A numpy array representing an image
+                - A PIL Image object
+
+        Returns:
+            `dict` with "path" and "bytes" fields for each image in the sequence
+        """
+        return [super().encode_example(img) for img in value]
+
+    def decode_example(
+        self, value: dict, token_per_repo_id=None
+    ) -> list["PIL.Image.Image"]:
+        """Decode example CBZ file into image data.
+
+        Args:
+            value (`str` or `dict`):
+                Either a string with absolute path to CBZ file, or a dictionary
+                with keys:
+                - `path`: String with absolute path to CBZ file.
+                - `bytes`: The bytes of the CBZ file.
+            token_per_repo_id (`dict`, *optional*):
+                To access and decode files from private repositories on the Hub.
+
+        Returns:
+            `List[PIL.Image.Image]`: List of images from CBZ
+        """
+        if isinstance(value, str):
+            zip_file = value
+        else:
+            zip_file = (
+                io.BytesIO(value["bytes"]) if value.get("bytes") else value["path"]
+            )
+
+        with zipfile.ZipFile(zip_file, "r") as zip_ref:
+            return [
+                super().decode_example(img_file)
+                for img_file in sorted(zip_ref.namelist())
+            ]
+
+
+class NhentaiConfig(datasets.folder_based_builder.FolderBasedBuilderConfig):
+    """BuilderConfig for ImageFolder."""
+
+    drop_labels: bool = None
+    drop_metadata: bool = None
+
+    def __post_init__(self):
+        super().__post_init__()
+
+
+class Nhentai(datasets.folder_based_builder.FolderBasedBuilder):
+    BASE_COLUMN_NAME = "comic"
+    BASE_FEATURE = Comic
+    BUILDER_CONFIG_CLASS = NhentaiConfig
+    BUILDER_CONFIGS = [
+        NhentaiConfig(
+            name="nhentai",
+            description="Default configuration for nhentai dataset",
+            drop_labels=False,
+            drop_metadata=False,
+        )
+    ]
+    DEFAULT_CONFIG_NAME = "nhentai"
+    EXTENSIONS: list[str] = [".cbz"]
+    VERSION = datasets.Version("1.0.0")
+
+    def _info(self):
+        """Returns the dataset metadata."""
+        return datasets.DatasetInfo(
+            description="Dataset of comic books in CBZ format",
+            features=datasets.Features(
+                {
+                    "comic": Comic(),
+                    "title": datasets.Value("string"),
+                    "media_id": datasets.Value("int64"),
+                    "num_favorites": datasets.Value("int64"),
+                    "tag": datasets.Sequence(datasets.Value("string")),
+                    "language": datasets.Sequence(datasets.Value("string")),
+                    "artist": datasets.Sequence(datasets.Value("string")),
+                    "category": datasets.Sequence(datasets.Value("string")),
+                    "num_pages": datasets.Value("int64"),
+                    "scanlator": datasets.Value("string"),
+                    "group": datasets.Sequence(datasets.Value("string")),
+                    "parody": datasets.Sequence(datasets.Value("string")),
+                    "character": datasets.Sequence(datasets.Value("string")),
+                    "epos": datasets.Value("int64"),
+                    "file_name": datasets.Value("string"),
+                }
+            ),
+            supervised_keys=None,
+            homepage="https://nhentai.net/",
+        )