Removing all Optional instances.

src/facebias/__init__.py

@@ -5,7 +5,7 @@ import logging
 from collections import OrderedDict
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Any, Callable, Optional
+from typing import Any, Callable
 
 import cv2
 import numpy as np
@@ -41,9 +41,9 @@ def load_metadata(p: Path, key_id="image", key_proc_fn=None) -> dict[str, dict[s
 
 def load_dataset(
         root: Path,
-        meta_path: Optional[Path] = None,
-        imname_proc_fn: Optional[Callable]=None
-) -> tuple[dict[Path, np.ndarray], Optional[dict[str, dict[str, Any]]]]:
+        meta_path: Path | None,
+        imname_proc_fn: Callable |None
+) -> tuple[dict[str, np.ndarray], dict[str, dict[str, Any]] | None]:
     """
     if `meta_path` is `None`, we won't attempt to read it.
 
@@ -53,20 +53,22 @@ def load_dataset(
         Root path to the images. We assume that the images are positioned in a
         flat directory under `root`.
 
-    meta_path: Optional[Path]
+    meta_path: Path | None
         Path to the metadata file on the dataset. Default is `None`.
 
-    imname_proc_fn: Optional[Callable]
+    imname_proc_fn: Callable | None
         Function to apply to the image filenames when building the output
         dictionary. Use it to uniformize it with the metadata file. Default
         is `None`.
 
     Returns
     -------
-    data: dict[Path, np.ndarray]
-    metadata, Optional[dict[str, dict[str, Any]]]
+    data: dict[str, np.ndarray]
+    metadata, dict[str, dict[str, Any]] | None
+        If metadata was not found, or not set to load (by leaving `meta_path`
+        as `None`), then we simply return `None`.
     """
-    metadata = dict()
+    metadata = None
     paths = set([p for p in root.iterdir() if not p.is_dir()])
     if meta_path is not None and meta_path in paths:
         # Just read the metadata and avoid needing to test for it in the main loop.
@@ -74,14 +76,13 @@ def load_dataset(
         paths.remove(meta_path)
 
     ims = OrderedDict()
-    # TODO(gschardong): Paralellize this?
     for p in paths:
         try:
             im = cv2.cvtColor(cv2.imread(p), cv2.COLOR_BGR2RGB)
         except cv2.error:
             logger.info(f'File "{p}" is not an image. Skipping.')
         else:
-            proc_imname = imname_proc_fn(p.name) if imname_proc_fn is not None else p.name
+            proc_imname = imname_proc_fn(p.name) if imname_proc_fn is not None else str(p.name)
             ims[proc_imname] = im
 
     if not metadata:

src/facebias/detectors/mediapipe.py

@@ -15,7 +15,7 @@ class MediapipeDetector(BaseFaceDetector):
     Parameters
     ----------
     model_path: str, PathLike
-    mode: Optional[mediapipe.tasks.python.vision.RunningMode]
+    mode: mediapipe.tasks.python.vision.RunningMode
     """
 
     def __init__(self, model_path: str, mode=vision.RunningMode.IMAGE) -> None:

src/facebias/estimators/fairface.py

@@ -2,7 +2,7 @@
 
 from collections import OrderedDict
 from pathlib import Path
-from typing import Any, Optional
+from typing import Any
 
 import numpy as np
 import torch
@@ -24,10 +24,10 @@ class FairFace(BaseEstimator):
 
     Parameters
     ----------
-    checkpoint_path: Optional[Path]
+    checkpoint_path: Path | None
         Path to the model weights stored as a PyTorch file.
 
-    device: Optional[torch.device]
+    device: torch.device
         The device to load the model into. By default is `"cpu"`
 
     Reference
@@ -36,8 +36,8 @@ class FairFace(BaseEstimator):
     """
     def __init__(
             self,
-            checkpoint_path: Optional[Path]=None,
-            device: Optional[torch.device]=torch.device("cpu")
+            checkpoint_path: Path | None,
+            device: torch.device=torch.device("cpu")
     ):
         self.T = transforms.Compose([
             transforms.ToPILImage(),

src/facebias/estimators/mivolov1.py

@@ -2,7 +2,7 @@
 
 from collections import OrderedDict
 from pathlib import Path
-from typing import Any, Optional
+from typing import Any
 
 import numpy as np
 import torch
@@ -24,7 +24,7 @@ class MiVOLOv1(BaseEstimator):
         Path to the saved checkpoint. Note that we expect the checkpoint to be
         in PyTorch format.
 
-    device: Optional[torch.device]
+    device: torch.device
         Device to load the checkpoints into. By default is "cpu".
 
     Reference
@@ -34,7 +34,7 @@ class MiVOLOv1(BaseEstimator):
     def __init__(
             self,
             checkpoint_path: Path,
-            device: Optional[torch.device]=torch.device("cpu")
+            device: torch.device=torch.device("cpu")
     ):
         self.T = transforms.Compose([
             transforms.ToPILImage(),