Speed up Image Enumeration (#17)

* feat: speed up dataset enumeration

Rather than attempting to load every image, now by default
we check if the file extention is known by Pillow the image
libary. The enhanced (checks can load each image) can be
utilised still by passing enhanced_validation=True to
`enumerate_images()`.

* style: lint fixes

* style: readability

camtrapml/dataset.py

@@ -2,9 +2,10 @@
 Dataset Handling
 """
 
+from typing import Union
 from pathlib import Path
 from os import walk
-from typing import Union
+from PIL import Image
 from camtrapml.image.utils import is_image
 
 
@@ -25,14 +26,28 @@ def __init__(self, path: Path, name: Union[None, str] = None):
         if not self.path.exists() or not self.path.is_dir():
             raise ValueError(f"{self.path} is not a directory")
 
-    def enumerate_images(self):
+    def enumerate_images(self, enhanced_validation: bool = False):
         """
         Enumerates all images in the dataset.
         """
+
+        supported_extensions = {
+            extention.lower() for extention, image_format in
+            Image.registered_extensions().items()
+            if image_format in Image.OPEN
+        }
+
         for root, _, files in walk(self.path):
             for file in files:
-                if is_image(Path(root) / file):
-                    yield Path(root) / file
+                file_path = Path(root) / file
+
+                if enhanced_validation:
+                    if is_image(file_path):
+                        yield file_path
+
+                else:
+                    if file_path.suffix.lower() in supported_extensions:
+                        yield file_path
 
     @staticmethod
     def from_coco(source):