add prototype image folder dataset (#4441)

* add prototype image folder dataset

* remove decoder datapipe

* [PROPOSAL] add RandomPicker

* refactor data loading

* fix mypy

* remove per-category datapipes

* fix mypy

Co-authored-by: Francisco Massa <[email protected]>

Author: pmeier

torchvision/prototype/__init__.py

@@ -0,0 +1 @@
+from . import datasets

torchvision/prototype/datasets/__init__.py

@@ -0,0 +1,2 @@
+from . import decoder
+from ._folder import from_data_folder, from_image_folder

torchvision/prototype/datasets/_folder.py

@@ -0,0 +1,81 @@
+import io
+import os
+import os.path
+import pathlib
+from typing import Callable, Optional, Collection
+from typing import Union, Tuple, List, Dict, Any
+
+import torch
+from torch.utils.data import IterDataPipe
+from torch.utils.data.datapipes.iter import FileLister, FileLoader, Mapper, Shuffler, Filter
+
+from torchvision.prototype.datasets.decoder import pil
+
+
+__all__ = ["from_data_folder", "from_image_folder"]
+
+# pseudo-infinite buffer size until a true infinite buffer is supported
+INFINITE = 1_000_000_000
+
+
+def _is_not_top_level_file(path: str, *, root: pathlib.Path) -> bool:
+    rel_path = pathlib.Path(path).relative_to(root)
+    return rel_path.is_dir() or rel_path.parent != pathlib.Path(".")
+
+
+def _collate_and_decode_data(
+    data: Tuple[str, io.IOBase],
+    *,
+    root: pathlib.Path,
+    categories: List[str],
+    decoder,
+) -> Dict[str, Any]:
+    path, buffer = data
+    data = decoder(buffer) if decoder else buffer
+    category = pathlib.Path(path).relative_to(root).parts[0]
+    label = torch.tensor(categories.index(category))
+    return dict(
+        path=path,
+        data=data,
+        label=label,
+        category=category,
+    )
+
+
+def from_data_folder(
+    root: Union[str, pathlib.Path],
+    *,
+    shuffler: Optional[Callable[[IterDataPipe], IterDataPipe]] = lambda dp: Shuffler(dp, buffer_size=INFINITE),
+    decoder: Optional[Callable[[io.IOBase], torch.Tensor]] = None,
+    valid_extensions: Optional[Collection[str]] = None,
+    recursive: bool = True,
+) -> Tuple[IterDataPipe, List[str]]:
+    root = pathlib.Path(root).expanduser().resolve()
+    categories = sorted(entry.name for entry in os.scandir(root) if entry.is_dir())
+    masks: Union[List[str], str] = [f"*.{ext}" for ext in valid_extensions] if valid_extensions is not None else ""
+    dp: IterDataPipe = FileLister(str(root), recursive=recursive, masks=masks)
+    dp = Filter(dp, _is_not_top_level_file, fn_kwargs=dict(root=root))
+    if shuffler:
+        dp = shuffler(dp)
+    dp = FileLoader(dp)
+    return (
+        Mapper(dp, _collate_and_decode_data, fn_kwargs=dict(root=root, categories=categories, decoder=decoder)),
+        categories,
+    )
+
+
+def _data_to_image_key(sample: Dict[str, Any]) -> Dict[str, Any]:
+    sample["image"] = sample.pop("data")
+    return sample
+
+
+def from_image_folder(
+    root: Union[str, pathlib.Path],
+    *,
+    decoder: Optional[Callable[[io.IOBase], torch.Tensor]] = pil,
+    valid_extensions: Collection[str] = ("jpg", "jpeg", "png", "ppm", "bmp", "pgm", "tif", "tiff", "webp"),
+    **kwargs: Any,
+) -> Tuple[IterDataPipe, List[str]]:
+    valid_extensions = [valid_extension for ext in valid_extensions for valid_extension in (ext.lower(), ext.upper())]
+    dp, categories = from_data_folder(root, decoder=decoder, valid_extensions=valid_extensions, **kwargs)
+    return Mapper(dp, _data_to_image_key), categories

torchvision/prototype/datasets/decoder.py

@@ -0,0 +1,12 @@
+import io
+
+import numpy as np
+import PIL.Image
+import torch
+
+__all__ = ["pil"]
+
+
+def pil(file: io.IOBase, mode="RGB") -> torch.Tensor:
+    image = PIL.Image.open(file).convert(mode.upper())
+    return torch.from_numpy(np.array(image, copy=True)).permute((2, 0, 1))