Cleanup: _image.py and _utils.py file in data subpackage (#1707)

* Add docstrings
* Add types

lightly/data/_image_loaders.py

@@ -1,6 +1,7 @@
-"""torchvision image loaders
-(see https://pytorch.org/docs/stable/_modules/torchvision/datasets/folder.html)
+"""Module for handling image loading in torchvision-compatible format.
 
+This module provides image loading functionality similar to torchvision's implementation
+(see https://pytorch.org/vision/main/generated/torchvision.datasets.ImageFolder.html)
 """
 
 # Copyright (c) 2020. Lightly AG and its affiliates.
@@ -9,15 +10,33 @@
 from PIL import Image
 
 
-def pil_loader(path):
+def pil_loader(path: str) -> Image.Image:
+    """Loads an image using PIL.
+
+    Args:
+        path: Path to the image file.
+
+    Returns:
+        A PIL Image in RGB format.
+    """
     # open path as file to avoid ResourceWarning
     # (https://github.com/python-pillow/Pillow/issues/835)
     with open(path, "rb") as f:
         img = Image.open(f)
         return img.convert("RGB")
 
 
-def accimage_loader(path):
+def accimage_loader(path: str) -> Image.Image:
+    """Loads an image using the accimage library for faster loading.
+
+    Falls back to PIL loader if accimage fails to load the image.
+
+    Args:
+        path: Path to the image file.
+
+    Returns:
+        An image loaded either by accimage or PIL in case of failure.
+    """
     try:
         import accimage
 
@@ -27,7 +46,17 @@ def accimage_loader(path):
         return pil_loader(path)
 
 
-def default_loader(path):
+def default_loader(path: str) -> Image.Image:
+    """Loads an image using the default backend specified in torchvision.
+
+    Uses accimage if available and configured as the backend, otherwise falls back to PIL.
+
+    Args:
+        path: Path to the image file.
+
+    Returns:
+        An image loaded by either accimage or PIL depending on the backend.
+    """
     from torchvision import get_image_backend
 
     if get_image_backend() == "accimage":

lightly/data/_utils.py

@@ -1,10 +1,15 @@
-""" Check for Corrupt Images """
+"""Provides functionality to identify corrupt images in a directory.
+
+This module helps users identify corrupt or unreadable image files within a specified
+directory. It uses parallel processing to efficiently scan through large collections
+of images.
+"""
 
 # Copyright (c) 2020. Lightly AG and its affiliates.
 # All Rights Reserved
 
 import os
-from typing import *
+from typing import List, Tuple
 
 import tqdm.contrib.concurrent as concurrent
 from PIL import Image, UnidentifiedImageError
@@ -13,18 +18,35 @@
 
 
 def check_images(data_dir: str) -> Tuple[List[str], List[str]]:
-    """Iterate through a directory of images and find corrupt images
+    """Identifies corrupt and healthy images in the specified directory.
+
+    The function attempts to open each image file in the directory to verify
+    its integrity. It processes images in parallel for better performance.
 
     Args:
-        data_dir: Path to the directory containing the images
+        data_dir: Directory path containing the image files to check.
 
     Returns:
-        (healthy_images, corrupt_images)
+        A tuple containing two lists:
+            - List of filenames of healthy images that can be opened successfully
+            - List of filenames of corrupt images that cannot be opened
+
+    Example:
+        >>> healthy, corrupt = check_images("path/to/images")
+        >>> print(f"Found {len(corrupt)} corrupt images")
     """
     dataset = LightlyDataset(input_dir=data_dir)
     filenames = dataset.get_filenames()
 
-    def _is_corrupt(filename):
+    def _is_corrupt(filename: str) -> bool:
+        """Checks if a single image file is corrupt.
+
+        Args:
+            filename: Name of the image file to check.
+
+        Returns:
+            True if the image is corrupt, False otherwise.
+        """
         try:
             image = Image.open(os.path.join(data_dir, filename))
             image.load()