dataeval 0.74.1__py3-none-any.whl → 0.75.0__py3-none-any.whl

dataeval/metrics/__init__.py

@@ -3,6 +3,6 @@ Metrics are a way to measure the performance of your models or datasets that
 can then be analyzed in the context of a given problem.
 """
 
-from dataeval.metrics import bias, estimators, stats
-
 __all__ = ["bias", "estimators", "stats"]
+
+from dataeval.metrics import bias, estimators, stats

dataeval/metrics/bias/__init__.py

@@ -3,22 +3,19 @@ Bias metrics check for skewed or imbalanced datasets and incomplete feature
 representation which may impact model performance.
 """
 
-from dataeval.metrics.bias.balance import BalanceOutput, balance
-from dataeval.metrics.bias.coverage import CoverageOutput, coverage
-from dataeval.metrics.bias.diversity import DiversityOutput, diversity
-from dataeval.metrics.bias.metadata_preprocessing import MetadataOutput, metadata_preprocessing
-from dataeval.metrics.bias.parity import ParityOutput, label_parity, parity
-
 __all__ = [
+    "BalanceOutput",
+    "CoverageOutput",
+    "DiversityOutput",
+    "ParityOutput",
     "balance",
     "coverage",
     "diversity",
     "label_parity",
     "parity",
-    "metadata_preprocessing",
-    "BalanceOutput",
-    "CoverageOutput",
-    "DiversityOutput",
-    "ParityOutput",
-    "MetadataOutput",
 ]
+
+from dataeval.metrics.bias.balance import BalanceOutput, balance
+from dataeval.metrics.bias.coverage import CoverageOutput, coverage
+from dataeval.metrics.bias.diversity import DiversityOutput, diversity
+from dataeval.metrics.bias.parity import ParityOutput, label_parity, parity

dataeval/metrics/bias/balance.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-__all__ = ["BalanceOutput", "balance"]
+__all__ = []
 
 import contextlib
 import warnings
@@ -12,9 +12,9 @@ import scipy as sp
 from numpy.typing import NDArray
 from sklearn.feature_selection import mutual_info_classif, mutual_info_regression
 
-from dataeval.metrics.bias.metadata_preprocessing import MetadataOutput
-from dataeval.metrics.bias.metadata_utils import get_counts, heatmap
 from dataeval.output import Output, set_metadata
+from dataeval.utils.metadata import Metadata, get_counts
+from dataeval.utils.plot import heatmap
 
 with contextlib.suppress(ImportError):
     from matplotlib.figure import Figure
@@ -119,7 +119,7 @@ def _validate_num_neighbors(num_neighbors: int) -> int:
 
 @set_metadata
 def balance(
-    metadata: MetadataOutput,
+    metadata: Metadata,
     num_neighbors: int = 5,
 ) -> BalanceOutput:
     """
@@ -127,14 +127,16 @@ def balance(
 
     Parameters
     ----------
-    metadata : MetadataOutput
-        Output after running `metadata_preprocessing`
+    metadata : Metadata
+        Preprocessed metadata from :func:`dataeval.utils.metadata.preprocess`
+    num_neighbors : int, default 5
+        Number of points to consider as neighbors
 
     Returns
     -------
     BalanceOutput
-        (num_factors+1) x (num_factors+1) estimate of mutual information
-        between num_factors metadata factors and class label. Symmetry is enforced.
+        (num_factors+1) x (num_factors+1) estimate of mutual information \
+            between num_factors metadata factors and class label. Symmetry is enforced.
 
     Note
     ----

dataeval/metrics/bias/coverage.py

@@ -1,18 +1,17 @@
 from __future__ import annotations
 
-__all__ = ["CoverageOutput", "coverage"]
+__all__ = []
 
 import contextlib
 import math
 from dataclasses import dataclass
-from typing import Literal
+from typing import Any, Literal
 
 import numpy as np
 from numpy.typing import ArrayLike, NDArray
 from scipy.spatial.distance import pdist, squareform
 
 from dataeval.interop import to_numpy
-from dataeval.metrics.bias.metadata_utils import coverage_plot
 from dataeval.output import Output, set_metadata
 from dataeval.utils.shared import flatten
 
@@ -20,6 +19,55 @@ with contextlib.suppress(ImportError):
     from matplotlib.figure import Figure
 
 
+def _plot(images: NDArray[Any], num_images: int) -> Figure:
+    """
+    Creates a single plot of all of the provided images
+
+    Parameters
+    ----------
+    images : NDArray
+        Array containing only the desired images to plot
+
+    Returns
+    -------
+    matplotlib.figure.Figure
+        Plot of all provided images
+    """
+    import matplotlib.pyplot as plt
+
+    num_images = min(num_images, len(images))
+
+    if images.ndim == 4:
+        images = np.moveaxis(images, 1, -1)
+    elif images.ndim == 3:
+        images = np.repeat(images[:, :, :, np.newaxis], 3, axis=-1)
+    else:
+        raise ValueError(
+            f"Expected a (N,C,H,W) or a (N, H, W) set of images, but got a {images.ndim}-dimensional set of images."
+        )
+
+    rows = int(np.ceil(num_images / 3))
+    fig, axs = plt.subplots(rows, 3, figsize=(9, 3 * rows))
+
+    if rows == 1:
+        for j in range(3):
+            if j >= len(images):
+                continue
+            axs[j].imshow(images[j])
+            axs[j].axis("off")
+    else:
+        for i in range(rows):
+            for j in range(3):
+                i_j = i * 3 + j
+                if i_j >= len(images):
+                    continue
+                axs[i, j].imshow(images[i_j])
+                axs[i, j].axis("off")
+
+    fig.tight_layout()
+    return fig
+
+
 @dataclass(frozen=True)
 class CoverageOutput(Output):
     """
@@ -62,7 +110,7 @@ class CoverageOutput(Output):
         selected_images = images[highest_uncovered_indices]
 
         # Plot the images
-        fig = coverage_plot(selected_images, top_k)
+        fig = _plot(selected_images, top_k)
 
         return fig
 

dataeval/metrics/bias/diversity.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-__all__ = ["DiversityOutput", "diversity"]
+__all__ = []
 
 import contextlib
 from dataclasses import dataclass
@@ -10,15 +10,44 @@ import numpy as np
 import scipy as sp
 from numpy.typing import ArrayLike, NDArray
 
-from dataeval.metrics.bias.metadata_preprocessing import MetadataOutput
-from dataeval.metrics.bias.metadata_utils import diversity_bar_plot, get_counts, heatmap
 from dataeval.output import Output, set_metadata
+from dataeval.utils.metadata import Metadata, get_counts
+from dataeval.utils.plot import heatmap
 from dataeval.utils.shared import get_method
 
 with contextlib.suppress(ImportError):
     from matplotlib.figure import Figure
 
 
+def _plot(labels: NDArray[Any], bar_heights: NDArray[Any]) -> Figure:
+    """
+    Plots a formatted bar plot
+
+    Parameters
+    ----------
+    labels : NDArray
+        Array containing the labels for each bar
+    bar_heights : NDArray
+        Array containing the values for each bar
+
+    Returns
+    -------
+    matplotlib.figure.Figure
+        Bar plot figure
+    """
+    import matplotlib.pyplot as plt
+
+    fig, ax = plt.subplots(figsize=(10, 10))
+
+    ax.bar(labels, bar_heights)
+    ax.set_xlabel("Factors")
+
+    plt.setp(ax.get_xticklabels(), rotation=45, ha="right", rotation_mode="anchor")
+
+    fig.tight_layout()
+    return fig
+
+
 @dataclass(frozen=True)
 class DiversityOutput(Output):
     """
@@ -77,8 +106,7 @@ class DiversityOutput(Output):
         else:
             # Creating label array for heat map axes
             heat_labels = np.concatenate((["class"], self.factor_names))
-
-            fig = diversity_bar_plot(heat_labels, self.diversity_index)
+            fig = _plot(heat_labels, self.diversity_index)
 
         return fig
 
@@ -165,7 +193,7 @@ def diversity_simpson(
 
 @set_metadata
 def diversity(
-    metadata: MetadataOutput,
+    metadata: Metadata,
     method: Literal["simpson", "shannon"] = "simpson",
 ) -> DiversityOutput:
     """
@@ -179,8 +207,8 @@ def diversity(
 
     Parameters
     ----------
-    metadata : MetadataOutput
-        Output after running `metadata_preprocessing`
+    metadata : Metadata
+        Preprocessed metadata from :func:`dataeval.utils.metadata.preprocess`
 
     Note
     ----
@@ -199,21 +227,21 @@ def diversity(
 
     >>> div_simp = diversity(metadata, method="simpson")
     >>> div_simp.diversity_index
-    array([0.72413793, 0.88636364, 0.72413793])
+    array([0.6       , 0.80882353, 1.        , 0.8       ])
 
     >>> div_simp.classwise
-    array([[0.69230769, 0.68965517],
-           [0.5       , 0.8       ]])
+    array([[0.5       , 0.8       , 0.8       ],
+           [0.63043478, 0.97560976, 0.52830189]])
 
     Compute Shannon diversity index of metadata and class labels
 
     >>> div_shan = diversity(metadata, method="shannon")
     >>> div_shan.diversity_index
-    array([0.8812909 , 0.96748876, 0.8812909 ])
+    array([0.81127812, 0.9426312 , 1.        , 0.91829583])
 
     >>> div_shan.classwise
-    array([[0.91651644, 0.86312057],
-           [0.68260619, 0.91829583]])
+    array([[0.68260619, 0.91829583, 0.91829583],
+           [0.81443569, 0.99107606, 0.76420451]])
 
     See Also
     --------

dataeval/metrics/bias/parity.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-__all__ = ["ParityOutput", "parity", "label_parity"]
+__all__ = []
 
 import warnings
 from dataclasses import dataclass
@@ -12,8 +12,8 @@ from scipy.stats import chisquare
 from scipy.stats.contingency import chi2_contingency, crosstab
 
 from dataeval.interop import as_numpy, to_numpy
-from dataeval.metrics.bias.metadata_preprocessing import MetadataOutput
 from dataeval.output import Output, set_metadata
+from dataeval.utils.metadata import Metadata
 
 TData = TypeVar("TData", np.float64, NDArray[np.float64])
 
@@ -167,8 +167,9 @@ def label_parity(
     --------
     Randomly creating some label distributions using ``np.random.default_rng``
 
-    >>> expected_labels = np_random_gen.choice([0, 1, 2, 3, 4], (100))
-    >>> observed_labels = np_random_gen.choice([2, 3, 0, 4, 1], (100))
+    >>> rng = np.random.default_rng(175)
+    >>> expected_labels = rng.choice([0, 1, 2, 3, 4], (100))
+    >>> observed_labels = rng.choice([2, 3, 0, 4, 1], (100))
     >>> label_parity(expected_labels, observed_labels)
     ParityOutput(score=14.007374204742625, p_value=0.0072715574616218, metadata_names=None)
     """
@@ -205,7 +206,7 @@ def label_parity(
 
 
 @set_metadata
-def parity(metadata: MetadataOutput) -> ParityOutput[NDArray[np.float64]]:
+def parity(metadata: Metadata) -> ParityOutput[NDArray[np.float64]]:
     """
     Calculate chi-square statistics to assess the linear relationship between multiple factors
     and class labels.
@@ -216,8 +217,8 @@ def parity(metadata: MetadataOutput) -> ParityOutput[NDArray[np.float64]]:
 
     Parameters
     ----------
-    metadata : MetadataOutput
-        Output after running `metadata_preprocessing`
+    metadata : Metadata
+        Preprocessed metadata from :func:`dataeval.utils.metadata.preprocess`
 
     Returns
     -------
@@ -249,16 +250,18 @@ def parity(metadata: MetadataOutput) -> ParityOutput[NDArray[np.float64]]:
     --------
     Randomly creating some "continuous" and categorical variables using ``np.random.default_rng``
 
-    >>> labels = np_random_gen.choice([0, 1, 2], (100))
+    >>> from dataeval.utils.metadata import preprocess
+    >>> rng = np.random.default_rng(175)
+    >>> labels = rng.choice([0, 1, 2], (100))
     >>> metadata_dict = [
     ...     {
-    ...         "age": list(np_random_gen.choice([25, 30, 35, 45], (100))),
-    ...         "income": list(np_random_gen.choice([50000, 65000, 80000], (100))),
-    ...         "gender": list(np_random_gen.choice(["M", "F"], (100))),
+    ...         "age": list(rng.choice([25, 30, 35, 45], (100))),
+    ...         "income": list(rng.choice([50000, 65000, 80000], (100))),
+    ...         "gender": list(rng.choice(["M", "F"], (100))),
     ...     }
     ... ]
     >>> continuous_factor_bincounts = {"age": 4, "income": 3}
-    >>> metadata = metadata_preprocessing(metadata_dict, labels, continuous_factor_bincounts)
+    >>> metadata = preprocess(metadata_dict, labels, continuous_factor_bincounts)
     >>> parity(metadata)
     ParityOutput(score=array([7.35731943, 5.46711299, 0.51506212]), p_value=array([0.28906231, 0.24263543, 0.77295762]), metadata_names=['age', 'income', 'gender'])
     """  # noqa: E501

dataeval/metrics/estimators/__init__.py

@@ -2,8 +2,8 @@
 Estimators calculate performance bounds and the statistical distance between datasets.
 """
 
+__all__ = ["ber", "divergence", "uap", "BEROutput", "DivergenceOutput", "UAPOutput"]
+
 from dataeval.metrics.estimators.ber import BEROutput, ber
 from dataeval.metrics.estimators.divergence import DivergenceOutput, divergence
 from dataeval.metrics.estimators.uap import UAPOutput, uap
-
-__all__ = ["ber", "divergence", "uap", "BEROutput", "DivergenceOutput", "UAPOutput"]

dataeval/metrics/estimators/ber.py

@@ -5,11 +5,12 @@ KNN based estimate for the :term:`Bayes error rate<Bayes Error Rate (BER)>`
 
 Learning to Bound the Multi-class Bayes Error (Th. 3 and Th. 4)
 https://arxiv.org/abs/1811.06419
+
 """
 
 from __future__ import annotations
 
-__all__ = ["BEROutput", "ber"]
+__all__ = []
 
 from dataclasses import dataclass
 from typing import Literal
@@ -38,6 +39,7 @@ class BEROutput(Output):
     """
 
     ber: float
+
     ber_lower: float
 
 

dataeval/metrics/estimators/divergence.py

@@ -5,7 +5,7 @@ using the Fast Nearest Neighbor and Minimum Spanning Tree algorithms
 
 from __future__ import annotations
 
-__all__ = ["DivergenceOutput", "divergence"]
+__all__ = []
 
 from dataclasses import dataclass
 from typing import Literal

dataeval/metrics/estimators/uap.py

@@ -6,7 +6,7 @@ average precision<Upper-Bound Average Precision (UAP)>` using empirical mean pre
 
 from __future__ import annotations
 
-__all__ = ["UAPOutput", "uap"]
+__all__ = []
 
 from dataclasses import dataclass
 

dataeval/metrics/stats/__init__.py

@@ -3,6 +3,24 @@ Statistics metrics calculate a variety of image properties and pixel statistics
 and label statistics against the images and labels of a dataset.
 """
 
+__all__ = [
+    "ChannelStatsOutput",
+    "DatasetStatsOutput",
+    "DimensionStatsOutput",
+    "HashStatsOutput",
+    "LabelStatsOutput",
+    "PixelStatsOutput",
+    "VisualStatsOutput",
+    "boxratiostats",
+    "channelstats",
+    "datasetstats",
+    "dimensionstats",
+    "hashstats",
+    "labelstats",
+    "pixelstats",
+    "visualstats",
+]
+
 from dataeval.metrics.stats.boxratiostats import boxratiostats
 from dataeval.metrics.stats.datasetstats import (
     ChannelStatsOutput,
@@ -15,21 +33,3 @@ from dataeval.metrics.stats.hashstats import HashStatsOutput, hashstats
 from dataeval.metrics.stats.labelstats import LabelStatsOutput, labelstats
 from dataeval.metrics.stats.pixelstats import PixelStatsOutput, pixelstats
 from dataeval.metrics.stats.visualstats import VisualStatsOutput, visualstats
-
-__all__ = [
-    "boxratiostats",
-    "channelstats",
-    "datasetstats",
-    "dimensionstats",
-    "hashstats",
-    "labelstats",
-    "pixelstats",
-    "visualstats",
-    "ChannelStatsOutput",
-    "DatasetStatsOutput",
-    "DimensionStatsOutput",
-    "HashStatsOutput",
-    "LabelStatsOutput",
-    "PixelStatsOutput",
-    "VisualStatsOutput",
-]

dataeval/metrics/stats/base.py

@@ -193,7 +193,7 @@ class StatsProcessorOutput(NamedTuple):
     results: list[dict[str, Any]]
     source_indices: list[SourceIndex]
     box_counts: list[int]
-    warnings_list: list[tuple[int, int, NDArray[np.float64], tuple[int, ...]]]
+    warnings_list: list[str]
 
 
 def process_stats(
@@ -206,13 +206,13 @@ def process_stats(
     results_list: list[dict[str, Any]] = []
     source_indices: list[SourceIndex] = []
     box_counts: list[int] = []
-    warnings_list: list[tuple[int, int, NDArray[np.float64], tuple[int, ...]]] = []
+    warnings_list: list[str] = []
     nboxes = [None] if boxes is None else normalize_box_shape(boxes)
     for i_b, box in enumerate(nboxes):
         i_b = None if box is None else i_b
         processor_list = [p(image, box, per_channel) for p in stats_processor_cls]
         if any(not p._is_valid_slice for p in processor_list) and i_b is not None and box is not None:
-            warnings_list.append((i, i_b, box, image.shape))
+            warnings_list.append(f"Bounding box [{i}][{i_b}]: {box} is out of bounds of {image.shape}.")
         results_list.append({k: v for p in processor_list for k, v in p.process().items()})
         if per_channel:
             source_indices.extend([SourceIndex(i, i_b, c) for c in range(image_boxes[0].shape[-3])])
@@ -302,7 +302,7 @@ def run_stats(
 
     # warnings are not emitted while in multiprocessing pools so we emit after gathering all warnings
     for w in warning_list:
-        warnings.warn(f"Bounding box [{w[0]}][{w[1]}]: {w[2]} is out of bounds of {w[3]}.", UserWarning)
+        warnings.warn(w, UserWarning)
 
     output = {}
     for results in results_list:

dataeval/metrics/stats/boxratiostats.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-__all__ = ["boxratiostats"]
+__all__ = []
 
 import copy
 from typing import Any, Callable, Generic, TypeVar, cast
@@ -130,17 +130,16 @@ def boxratiostats(
     --------
     Calculating the box ratio statistics using the dimension stats of the boxes and images
 
-    >>> imagestats = dimensionstats(images)
-    >>> boxstats = dimensionstats(images, bboxes)
+    >>> from dataeval.metrics.stats import dimensionstats
+    >>> imagestats = dimensionstats(stats_images)
+    >>> boxstats = dimensionstats(stats_images, bboxes)
     >>> ratiostats = boxratiostats(boxstats, imagestats)
     >>> print(ratiostats.aspect_ratio)
-    [ 1.15169271  0.78450521 21.33333333  1.5234375   2.25651042  0.77799479
-      0.88867188  3.40625     1.73307292  1.11132812  0.75018315  0.45018315
-      0.69596354 20.          5.11197917  2.33333333  0.75        0.70019531]
+    [ 0.86376953  0.58837891 16.          0.85714286  1.26959707  0.43772894
+      0.66650391  3.83296703  1.95018315]
     >>> print(ratiostats.size)
-    [0.03401693 0.01383464 0.00130208 0.01822917 0.02327474 0.00683594
-     0.01220703 0.0168457  0.01057943 0.00976562 0.00130208 0.01098633
-     0.02246094 0.0012207  0.01123047 0.00911458 0.02636719 0.06835938]
+    [0.0255127  0.01037598 0.00097656 0.01822917 0.02327474 0.00683594
+     0.00915527 0.03369141 0.02115885]
     """
     output_cls = type(boxstats)
     if type(boxstats) is not type(imgstats):

dataeval/metrics/stats/datasetstats.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-__all__ = ["DatasetStatsOutput", "ChannelStatsOutput", "datasetstats", "channelstats"]
+__all__ = []
 
 from dataclasses import dataclass
 from typing import Any, Iterable
@@ -25,7 +25,7 @@ class DatasetStatsOutput(Output):
 
     This class represents the outputs of various stats functions against a single
     dataset, such that each index across all stat outputs are representative of
-    the same source image.  Modifying or mixing outputs will result in inaccurate
+    the same source image. Modifying or mixing outputs will result in inaccurate
     outlier calculations if not created correctly.
 
     Attributes
@@ -60,7 +60,7 @@ class ChannelStatsOutput(Output):
 
     This class represents the outputs of various per-channel stats functions against
     a single dataset, such that each index across all stat outputs are representative
-    of the same source image.  Modifying or mixing outputs will result in inaccurate
+    of the same source image. Modifying or mixing outputs will result in inaccurate
     outlier calculations if not created correctly.
 
     Attributes
@@ -119,13 +119,11 @@ def datasetstats(
     --------
     Calculating the dimension, pixel and visual stats for a dataset with bounding boxes
 
-    >>> stats = datasetstats(images, bboxes)
+    >>> stats = datasetstats(stats_images, bboxes)
     >>> print(stats.dimensionstats.aspect_ratio)
-    [ 0.864   0.5884 16.      1.143   1.692   0.5835  0.6665  2.555   1.3
-      0.8335  1.      0.6     0.522  15.      3.834   1.75    0.75    0.7   ]
-    >>> print(stats.visualstats.contrast)
-    [1.744   1.946   0.1164  0.0635  0.0633  0.06274 0.0429  0.0317  0.0317
-     0.02576 0.02081 0.02171 0.01915 0.01767 0.01799 0.01595 0.01433 0.01478]
+    [ 0.864   0.5884 16.      1.143   1.692   0.5835  0.6665  2.555   1.3   ]
+    >>> print(stats.visualstats.sharpness)
+    [4.04   4.434  0.2778 4.957  5.145  5.22   4.957  3.076  2.855 ]
     """
     outputs = run_stats(images, bboxes, False, [DimensionStatsProcessor, PixelStatsProcessor, VisualStatsProcessor])
     return DatasetStatsOutput(*outputs, labelstats=labelstats(labels) if labels else None)  # type: ignore
@@ -162,12 +160,10 @@ def channelstats(
     --------
     Calculating the per-channel pixel and visual stats for a dataset
 
-    >>> stats = channelstats(images)
+    >>> stats = channelstats(stats_images)
     >>> print(stats.visualstats.darkness)
-    [0.07495 0.1748  0.275   0.1047  0.11096 0.1172  0.2047  0.2109  0.2172
-     0.3047  0.311   0.3171  0.4048  0.411   0.4172  0.505   0.5107  0.517
-     0.6045  0.611   0.617   0.7046  0.711   0.7173  0.8047  0.811   0.8174
-     0.905   0.911   0.917  ]
+    [0.1499 0.3499 0.55   0.2094 0.2219 0.2344 0.4194 0.6094 0.622  0.6343
+     0.8154]
     """
     outputs = run_stats(images, bboxes, True, [PixelStatsProcessor, VisualStatsProcessor])
     return ChannelStatsOutput(*outputs)  # type: ignore

dataeval/metrics/stats/dimensionstats.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-__all__ = ["DimensionStatsOutput", "dimensionstats"]
+__all__ = []
 
 from dataclasses import dataclass
 from typing import Any, Callable, Iterable
@@ -106,10 +106,10 @@ def dimensionstats(
     --------
     Calculating the dimension statistics on the images, whose shape is (C, H, W)
 
-    >>> results = dimensionstats(images)
+    >>> results = dimensionstats(stats_images)
     >>> print(results.aspect_ratio)
-    [0.75  0.75  0.75  0.75  0.75  0.75  1.333 0.75  0.75  1.   ]
+    [1.     1.     1.333  1.     0.6665]
     >>> print(results.channels)
-    [1 1 1 1 1 1 3 1 1 3]
+    [3 3 1 3 1]
     """
     return run_stats(images, bboxes, False, [DimensionStatsProcessor])[0]

dataeval/metrics/stats/hashstats.py

@@ -1,6 +1,8 @@
 from __future__ import annotations
 
-__all__ = ["HashStatsOutput", "hashstats"]
+import warnings
+
+__all__ = []
 
 from dataclasses import dataclass
 from typing import Callable, Iterable
@@ -41,7 +43,7 @@ def pchash(image: ArrayLike) -> str:
     """
     Performs a perceptual hash on an image by resizing to a square NxN image
     using the Lanczos algorithm where N is 32x32 or the largest multiple of
-    8 that is smaller than the input image dimensions.  The resampled image
+    8 that is smaller than the input image dimensions. The resampled image
     is compressed using a discrete cosine transform and the lowest frequency
     component is encoded as a bit array of greater or less than median value
     and returned as a hex string.
@@ -54,13 +56,15 @@ def pchash(image: ArrayLike) -> str:
     Returns
     -------
     str
-        The hex string hash of the image using perceptual hashing
+        The hex string hash of the image using perceptual hashing, or empty
+        string if the image is too small to be hashed
     """
     # Verify that the image is at least larger than an 8x8 image
     arr = as_numpy(image)
     min_dim = min(arr.shape[-2:])
     if min_dim < HASH_SIZE + 1:
-        raise ValueError(f"Image must be larger than {HASH_SIZE}x{HASH_SIZE} for fuzzy hashing.")
+        warnings.warn(f"Image must be larger than {HASH_SIZE}x{HASH_SIZE} for fuzzy hashing.")
+        return ""
 
     # Calculates the dimensions of the resized square image
     resize_dim = HASH_SIZE * min((min_dim - 1) // HASH_SIZE, MAX_FACTOR)
@@ -92,7 +96,7 @@ def pchash(image: ArrayLike) -> str:
 def xxhash(image: ArrayLike) -> str:
     """
     Performs a fast non-cryptographic hash using the xxhash algorithm
-    (xxhash.com) against the image as a flattened bytearray.  The hash
+    (xxhash.com) against the image as a flattened bytearray. The hash
     is returned as a hex string.
 
     Parameters
@@ -147,10 +151,10 @@ def hashstats(
     --------
     Calculating the statistics on the images, whose shape is (C, H, W)
 
-    >>> results = hashstats(images)
+    >>> results = hashstats(stats_images)
     >>> print(results.xxhash)
-    ['a72434443d6e7336', 'efc12c2f14581d79', '4a1e03483a27d674', '3a3ecedbcf814226']
+    ['6274f837b34ed9f0', '256504fdb6e3d2a4', '7dd0c56ca8474fb0', '50956ad4592f5bbc', '5ba2354079d42aa5']
     >>> print(results.pchash)
-    ['8f25506af46a7c6a', '8000808000008080', '8e71f18e0ef18e0e', 'a956d6a956d6a928']
+    ['a666999999666666', 'e666999999266666', 'e666999966663299', 'e666999999266666', '96e91656e91616e9']
     """
     return run_stats(images, bboxes, False, [HashStatsProcessor])[0]