edgefirst-validator 4.2.1__py3-none-any.whl

edgefirst/validator/metrics/data/plots.py

@@ -0,0 +1,476 @@
+import numpy as np
+
+
+class Plots:
+    """
+    Container used to store the data needed for
+    plotting the validation charts.
+
+    Parameters
+    ----------
+    labels: list
+        A list of unique string labels to
+        initialize the confusion matrix.
+    """
+
+    def __init__(self, labels: list = []):
+        self.labels = labels
+        self.reset()
+
+    @property
+    def class_histogram_data(self) -> dict:
+        """
+        Attribute to access the class histogram data.
+
+        Returns
+        -------
+        dict
+            This contains the data for the class histogram.
+        """
+        return self.__class_histogram_data
+
+    @class_histogram_data.setter
+    def class_histogram_data(self, data: dict):
+        """
+        Sets the data for the class histogram to a new value.
+
+        Parameters
+        ----------
+        data: dict
+            This is the class histogram data to set.
+            This should be a dictionary with the following keys.
+
+                .. code-block:: python
+
+                    {
+                        "precision": precision,
+                        "recall": recall,
+                        "accuracy": accuracy,
+                        "tp": tp,
+                        "fn": fn,
+                        "fp": fp,
+                        "gt": gt
+                    }
+        """
+        self.__class_histogram_data = data
+
+    def append_class_histogram_data(self, label: str, data: dict):
+        """
+        This adds another key to the class histogram data indicated as the
+        class label and data contains the metrics of that label.
+
+        Parameters
+        ----------
+        label: str
+            This is the key of the dictionary that is the class label.
+        data: dict
+            This contains the metrics of the label. This should
+            be a dictionary with the following keys.
+
+                .. code-block:: python
+
+                    {
+                        "precision": precision,
+                        "recall": recall,
+                        "accuracy": accuracy,
+                        "tp": tp,
+                        "fn": fn,
+                        "fp": fp,
+                        "gt": gt
+                    }
+        """
+        self.__class_histogram_data[label] = data
+
+    @property
+    def confusion_labels(self) -> list:
+        """
+        Attribute to access the confusion matrix unique labels.
+
+        Returns
+        -------
+        list
+            This contains the labels for the confusion matrix.
+        """
+        return self.__confusion_labels
+
+    @confusion_labels.setter
+    def confusion_labels(self, labels: list):
+        """
+        Sets the labels for the confusion matrix to a new value.
+
+        Parameters
+        ----------
+        labels: list
+            These are the confusion matrix labels to set.
+        """
+        self.__confusion_labels = labels
+
+    @property
+    def confusion_matrix(self) -> np.ndarray:
+        """
+        Attribute to access the confusion matrix.
+
+        Returns
+        -------
+        np.ndarray
+            This contains the confusion matrix.
+        """
+        return self.__confusion_matrix
+
+    @confusion_matrix.setter
+    def confusion_matrix(self, matrix: np.ndarray):
+        """
+        Sets the confusion matrix to a new value.
+
+        Parameters
+        ----------
+        matrix: :py:class:`np.ndarray`
+            This is the confusion matrix to set.
+        """
+        self.__confusion_matrix = matrix
+
+    def initialize_confusion_matrix(self):
+        """
+        Initialize the confusion matrix array.
+        """
+        labels = self.labels
+        # Insert the background class in the confusion matrix if it doesn't
+        # exist.
+        if not any(s.lower() == "background" for s in self.labels):
+            labels = ["background"] + self.labels
+
+        self.confusion_labels = labels
+        # Rows = predictions, columns = ground truth
+        self.confusion_matrix = np.zeros(
+            (len(labels), len(labels)), dtype=np.int32)
+
+    @property
+    def px(self) -> np.ndarray:
+        """
+        Attribute to access px.
+
+        Returns
+        -------
+        np.ndarray
+            Precision vs Recall Curve 1000-point interpolated px values
+            representing recall.
+        """
+        return self.__px
+
+    @px.setter
+    def px(self, this_px: np.ndarray):
+        """
+        Sets px to a new value.
+
+        Parameters
+        ----------
+        px: np.ndarray
+            The px values to set.
+        """
+        self.__px = this_px
+
+    @property
+    def py(self) -> np.ndarray:
+        """
+        Attribute to access py.
+
+        Returns
+        -------
+        np.ndarray
+            Precision vs Recall Curve 1000-point interpolated py values
+            representing precision.
+        """
+        return self.__py
+
+    @py.setter
+    def py(self, this_py: np.ndarray):
+        """
+        Sets py to a new value.
+
+        Parameters
+        ----------
+        py: np.ndarray
+            The py values to set.
+        """
+        self.__py = this_py
+
+    @property
+    def precision(self) -> np.ndarray:
+        """
+        Attribute to access the array of precision values.
+
+        Returns
+        -------
+        np.ndarray
+            This contains the data for the precision recall curve.
+        """
+        return self.__precision
+
+    @precision.setter
+    def precision(self, data: np.ndarray):
+        """
+        Sets the data for the precision values.
+
+        Parameters
+        ----------
+        data: :py:class:`np.ndarray`
+            These are the precision values to set.
+
+            This data should be formatted as the following:
+            (nc x thresholds) so each row are for a unique class and
+            each column is the precision value for each score threshold.
+        """
+        self.__precision = data
+
+    @property
+    def recall(self) -> np.ndarray:
+        """
+        Attribute to access the array of recall values.
+
+        Returns
+        -------
+        np.ndarray
+            This contains the data for the precision recall curve.
+        """
+        return self.__recall
+
+    @recall.setter
+    def recall(self, data: np.ndarray):
+        """
+        Sets the data for the recall values.
+
+        Parameters
+        ----------
+        data: np.ndarray
+            These are the recall values to set.
+
+            This data should be formatted as the following:
+            (nc x thresholds) so each row are for a unique class and
+            each column is the recall value for each score threshold.
+        """
+        self.__recall = data
+
+    @property
+    def f1(self) -> np.ndarray:
+        """
+        Attribute to access the array of F1 values.
+
+        Returns
+        -------
+        np.ndarray
+            This contains the data for the precision-f1 curve.
+        """
+        return self.__f1
+
+    @f1.setter
+    def f1(self, data: np.ndarray):
+        """
+        Sets the data for the F1 values.
+
+        Parameters
+        ----------
+        data: np.ndarray
+            These are the F1 values to set.
+
+            This data should be formatted as the following:
+            (nc x thresholds) so each row are for a unique class and
+            each column is the F1 value for each score threshold.
+        """
+        self.__f1 = data
+
+    @property
+    def average_precision(self) -> np.ndarray:
+        """
+        Attribute to access the array of average precision values.
+
+        Returns
+        -------
+        np.ndarray
+            This contains the data for the precision recall curve.
+        """
+        return self.__average_precision
+
+    @average_precision.setter
+    def average_precision(self, data: np.ndarray):
+        """
+        Sets the data for the average precision values.
+
+        Parameters
+        ----------
+        data: np.ndarray
+            These are the average precision values to set.
+
+            This data should be formatted as the following:
+            (nc x 10) so each row are for a unique class and
+            each column is the precision at 10 different IoU threshold from
+            0.50 to 0.95 in 0.05 intervals with a static score threshold
+            set from the command line.
+        """
+        self.__average_precision = data
+
+    @property
+    def curve_labels(self) -> list:
+        """
+        Attribute to access the precision recall curve unique labels.
+
+        Returns
+        -------
+        list
+            This contains the labels for the precision recall curve.
+        """
+        return self.__curve_labels
+
+    @curve_labels.setter
+    def curve_labels(self, labels: list):
+        """
+        Sets the labels for the precision recall curve to a new value.
+
+        Parameters
+        ----------
+        labels: list
+            These are the precision recall curve labels to set.
+        """
+        self.__curve_labels = labels
+
+    @property
+    def tp_scores(self) -> list:
+        """
+        Attribute to access the confidence scores of true positive detections.
+
+        Returns
+        -------
+        list
+            A list containing the scores assigned to true positive detections.
+        """
+        return self.__tp_scores
+
+    @tp_scores.setter
+    def tp_scores(self, scores: list):
+        """
+        Sets the confidence scores for true positive detections.
+
+        Parameters
+        ----------
+        scores: list
+            A list of confidence scores to assign to true positive detections.
+        """
+        self.__tp_scores = scores
+
+    @property
+    def fp_scores(self) -> list:
+        """
+        Attribute to access the confidence scores of false positive detections.
+
+        Returns
+        -------
+        list
+            A list containing the scores assigned to false positive detections.
+        """
+        return self.__fp_scores
+
+    @fp_scores.setter
+    def fp_scores(self, scores: list):
+        """
+        Sets the confidence scores for false positive detections.
+
+        Parameters
+        ----------
+        scores: list
+            A list of confidence scores to assign to false positive detections.
+        """
+        self.__fp_scores = scores
+
+    @property
+    def tp_ious(self):
+        """
+        Attribute to access the IoU values for true positive detections.
+
+        Returns
+        -------
+        list
+            A list containing Intersection-over-Union (IoU)
+            values for true positives.
+        """
+        return self.__tp_ious
+
+    @tp_ious.setter
+    def tp_ious(self, ious: list):
+        """
+        Sets the IoU values for true positive detections.
+
+        Parameters
+        ----------
+        ious: list
+            A list of IoU values to assign to true positive detections.
+        """
+        self.__tp_ious = ious
+
+    @property
+    def fp_ious(self):
+        """
+        Attribute to access the IoU values for false positive detections.
+
+        Returns
+        -------
+        list
+            A list containing Intersection-over-Union (IoU) values
+            for false positives.
+        """
+        return self.__fp_ious
+
+    @fp_ious.setter
+    def fp_ious(self, ious: list):
+        """
+        Sets the IoU values for false positive detections.
+
+        Parameters
+        ----------
+        ious: list
+            A list of IoU values to assign to false positive detections.
+        """
+        self.__fp_ious = ious
+
+    def reset(self):
+        """
+        Resets the containers for the data use to plot.
+        """
+        self.__class_histogram_data = dict()
+
+        """Confusion Matrix Data"""
+        self.__confusion_labels = list()
+        self.__confusion_matrix = list()
+
+        """Precision Recall Curve"""
+        self.__px = np.array([])
+        self.__py = np.array([])
+        self.__precision = list()
+        self.__recall = list()
+        self.__f1 = list()
+        self.__average_precision = list()
+        self.__curve_labels = list()
+
+        """TP vs FP scores"""
+        self.__tp_scores = list()
+        self.__fp_scores = list()
+
+        """TP vs FP IoUs"""
+        self.__tp_ious = list()
+        self.__fp_ious = list()
+
+
+class MultitaskPlots:
+    """
+    A container for both detection and segmentation
+    plots for Multitask validation.
+
+    Parameters
+    -----------
+    detection_plots: Plots
+        Detection plots container.
+    segmentation_plots: Plots
+        Segmentation plots container.
+    """
+
+    def __init__(self, detection_plots: Plots, segmentation_plots: Plots):
+        self.detection_plots = detection_plots
+        self.segmentation_plots = segmentation_plots