Upload anime_segmentation.py

anime_segmentation.py

@@ -0,0 +1,76 @@
+import os
+import datasets
+import PIL
+from datasets import DownloadManager, DatasetInfo, load_dataset
+
+_DESCRIPTION = """\
+A segmentation dataset for anime character
+"""
+_HOMEPAGE = "https://huggingface.co/datasets/skytnt/anime-segmentation"
+_URL_BASE = "https://huggingface.co/datasets/skytnt/anime-segmentation/resolve/main/data/"
+
+
+class AnimeSegmentationConfig(datasets.BuilderConfig):
+
+    def __init__(self, features, data_files, **kwargs):
+        super(AnimeSegmentationConfig, self).__init__(**kwargs)
+        self.features = features
+        self.data_files = data_files
+
+
+class AnimeSegmentation(datasets.GeneratorBasedBuilder):
+    BUILDER_CONFIGS = [
+        AnimeSegmentationConfig(
+            name="bg",
+            description="background",
+            features=["image"],
+            data_files=["bg-00.zip", "bg-01.zip", "bg-02.zip", "bg-03.zip", "bg-04.zip"]
+        ),
+        AnimeSegmentationConfig(
+            name="fg",
+            description="foreground",
+            features=["image"],
+            data_files=["fg-00.zip", "fg-01.zip", "fg-02.zip", "fg-03.zip", "fg-04.zip", "fg-05.zip"]
+        ),
+        AnimeSegmentationConfig(
+            name="imgs-masks",
+            description="real images and masks",
+            features=["image", "mask"],
+            data_files=["imgs-masks.zip"]
+        )
+    ]
+
+    def _info(self) -> DatasetInfo:
+
+        features = {feature: datasets.Image() for feature in self.config.features}
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features(features),
+            supervised_keys=None,
+            homepage=_HOMEPAGE,
+            citation="",
+        )
+
+    def _split_generators(self, dl_manager: DownloadManager):
+        urls = [_URL_BASE + data_file for data_file in self.config.data_files]
+        dirs = dl_manager.download_and_extract(urls)
+        return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"dirs": dirs})]
+
+    def _generate_examples(self, dirs):
+        if self.config.name != "imgs-masks":
+            for path in dirs:
+                all_fnames = {os.path.relpath(os.path.join(root, fname), start=path)
+                              for root, _dirs, files in os.walk(path) for fname in files}
+                image_fnames = sorted(fname for fname in all_fnames
+                                      if os.path.splitext(fname)[1].lower() in PIL.Image.EXTENSION)
+                for image_fname in image_fnames:
+                    yield image_fname, {"image": os.path.join(path, image_fname)}
+        else:
+            path = dirs[0]
+            all_fnames = {os.path.relpath(os.path.join(root, fname), start=path)
+                          for root, _dirs, files in os.walk(path) for fname in files}
+            image_fnames = sorted(fname for fname in all_fnames if fname.startswith(
+                "imgs") and os.path.splitext(fname)[1].lower() in PIL.Image.EXTENSION)
+            for image_fname in image_fnames:
+                yield image_fname, {"image": os.path.join(path, image_fname),
+                                    "mask": os.path.join(path, image_fname.replace("imgs", "masks"))}