valor-lite 0.36.4__tar.gz → 0.36.6__tar.gz

{valor_lite-0.36.4 → valor_lite-0.36.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: valor-lite
-Version: 0.36.4
+Version: 0.36.6
 Summary: Evaluate machine learning models.
 Project-URL: homepage, https://www.striveworks.com
 Requires-Python: >=3.10

{valor_lite-0.36.4 → valor_lite-0.36.6}/valor_lite/object_detection/computation.py

@@ -345,12 +345,6 @@ def rank_pairs(
     mask_unmatched_predictions = ~np.isin(pairs[:, 2], matched_predictions)
     pairs = pairs[mask_label_match | mask_unmatched_predictions]
 
-    # remove predictions for labels that have no ground truths
-    for label_idx, count in enumerate(label_metadata[:, 0]):
-        if count > 0:
-            continue
-        pairs = pairs[pairs[:, 4] != label_idx]
-
     # only keep the highest ranked pair
     _, indices = np.unique(pairs[:, [0, 2, 4]], axis=0, return_index=True)
     pairs = pairs[indices]

{valor_lite-0.36.4 → valor_lite-0.36.6}/valor_lite/object_detection/manager.py

@@ -410,7 +410,6 @@ class Evaluator:
         )
         return unpack_precision_recall_into_metric_lists(
             results=results,
-            label_metadata=label_metadata,
             iou_thresholds=iou_thresholds,
             score_thresholds=score_thresholds,
             index_to_label=self.index_to_label,

{valor_lite-0.36.4 → valor_lite-0.36.6}/valor_lite/object_detection/utilities.py

@@ -23,7 +23,6 @@ def unpack_precision_recall_into_metric_lists(
     iou_thresholds: list[float],
     score_thresholds: list[float],
     index_to_label: list[str],
-    label_metadata: NDArray[np.int32],
 ):
     (
         (
@@ -48,7 +47,6 @@ def unpack_precision_recall_into_metric_lists(
         )
         for iou_idx, iou_threshold in enumerate(iou_thresholds)
         for label_idx, label in enumerate(index_to_label)
-        if int(label_metadata[label_idx, 0]) > 0
     ]
 
     metrics[MetricType.mAP] = [
@@ -67,7 +65,6 @@ def unpack_precision_recall_into_metric_lists(
             label=label,
         )
         for label_idx, label in enumerate(index_to_label)
-        if int(label_metadata[label_idx, 0]) > 0
     ]
 
     # TODO - (c.zaloom) will be removed in the future
@@ -87,7 +84,6 @@ def unpack_precision_recall_into_metric_lists(
         )
         for score_idx, score_threshold in enumerate(score_thresholds)
         for label_idx, label in enumerate(index_to_label)
-        if int(label_metadata[label_idx, 0]) > 0
     ]
 
     metrics[MetricType.mAR] = [
@@ -108,7 +104,6 @@ def unpack_precision_recall_into_metric_lists(
             label=label,
         )
         for label_idx, label in enumerate(index_to_label)
-        if int(label_metadata[label_idx, 0]) > 0
     ]
 
     # TODO - (c.zaloom) will be removed in the future
@@ -129,13 +124,9 @@ def unpack_precision_recall_into_metric_lists(
         )
         for iou_idx, iou_threshold in enumerate(iou_thresholds)
         for label_idx, label in enumerate(index_to_label)
-        if label_metadata[label_idx, 0] > 0
     ]
 
     for label_idx, label in enumerate(index_to_label):
-        if label_metadata[label_idx, 0] == 0:
-            continue
-
         for score_idx, score_threshold in enumerate(score_thresholds):
             for iou_idx, iou_threshold in enumerate(iou_thresholds):
 

{valor_lite-0.36.4 → valor_lite-0.36.6}/valor_lite/semantic_segmentation/annotation.py

@@ -1,3 +1,4 @@
+import warnings
 from dataclasses import dataclass, field
 
 import numpy as np
@@ -79,30 +80,37 @@ class Segmentation:
             )
         self.size = self.shape[0] * self.shape[1]
 
-        mask_accumulation = None
-        for groundtruth in self.groundtruths:
-            if self.shape != groundtruth.mask.shape:
-                raise ValueError(
-                    f"ground truth masks for datum '{self.uid}' should have shape '{self.shape}'. Received mask with shape '{groundtruth.mask.shape}'"
-                )
-
-            if mask_accumulation is None:
-                mask_accumulation = groundtruth.mask.copy()
-            elif np.logical_and(mask_accumulation, groundtruth.mask).any():
-                raise ValueError("ground truth masks cannot overlap")
-            else:
-                mask_accumulation = mask_accumulation | groundtruth.mask
+        self._validate_bitmasks(self.groundtruths, "ground truth")
+        self._validate_bitmasks(self.predictions, "prediction")
 
+    def _validate_bitmasks(self, bitmasks: list[Bitmask], key: str):
         mask_accumulation = None
-        for prediction in self.predictions:
-            if self.shape != prediction.mask.shape:
+        mask_overlap_accumulation = None
+        for idx, bitmask in enumerate(bitmasks):
+            if not isinstance(bitmask, Bitmask):
+                raise ValueError(f"expected 'Bitmask', got '{bitmask}'")
+            if self.shape != bitmask.mask.shape:
                 raise ValueError(
-                    f"prediction masks for datum '{self.uid}' should have shape '{self.shape}'. Received mask with shape '{prediction.mask.shape}'"
+                    f"{key} masks for datum '{self.uid}' should have shape '{self.shape}'. Received mask with shape '{bitmask.mask.shape}'"
                 )
 
             if mask_accumulation is None:
-                mask_accumulation = prediction.mask.copy()
-            elif np.logical_and(mask_accumulation, prediction.mask).any():
-                raise ValueError("prediction masks cannot overlap")
+                mask_accumulation = bitmask.mask.copy()
+                mask_overlap_accumulation = np.zeros_like(mask_accumulation)
+            elif np.logical_and(mask_accumulation, bitmask.mask).any():
+                mask_overlap = np.logical_and(mask_accumulation, bitmask.mask)
+                bitmasks[idx].mask[mask_overlap] = False
+                mask_overlap_accumulation = (
+                    mask_overlap_accumulation | mask_overlap
+                )
             else:
-                mask_accumulation = mask_accumulation | prediction.mask
+                mask_accumulation = mask_accumulation | bitmask.mask
+        if (
+            mask_overlap_accumulation is not None
+            and mask_overlap_accumulation.any()
+        ):
+            count = mask_overlap_accumulation.sum()
+            total = mask_overlap_accumulation.size
+            warnings.warn(
+                f"{key} masks for datum '{self.uid}' had {count} / {total} pixels overlapped."
+            )

{valor_lite-0.36.4 → valor_lite-0.36.6}/valor_lite.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: valor-lite
-Version: 0.36.4
+Version: 0.36.6
 Summary: Evaluate machine learning models.
 Project-URL: homepage, https://www.striveworks.com
 Requires-Python: >=3.10