clean API interface; working on tests for iou_type SEGM

Author: gianscarpe

tests/detection/test_map.py

@@ -266,6 +266,26 @@ def test_empty_metric():
     metric.compute()
 
 
+@pytest.mark.skipif(_pytest_condition, reason="test requires that torchvision=>0.8.0 is installed")
+def test_segm_iou_empty_mask():
+    """Test empty ground truths."""
+    metric = MeanAveragePrecision(iou_type="segm")
+
+    metric.update(
+        [
+            dict(
+                masks=torch.randint(0, 1, (1, 10, 10)),
+                scores=torch.Tensor([0.5]),
+                labels=torch.IntTensor([4]),
+            ),
+        ],
+        [
+            dict(masks=torch.Tensor([]), labels=torch.IntTensor([])),
+        ],
+    )
+    metric.compute()
+
+
 @pytest.mark.skipif(_pytest_condition, reason="test requires that torchvision=>0.8.0 is installed")
 def test_error_on_wrong_input():
     """Test class input validation."""

torchmetrics/detection/map.py

@@ -246,13 +246,11 @@ class MeanAveragePrecision(Metric):
             If ``class_metrics`` is not a boolean
     """
 
-    detection_boxes: List[Tensor]
+    detections: List[Tensor]
     detection_scores: List[Tensor]
     detection_labels: List[Tensor]
-    groundtruth_boxes: List[Tensor]
+    groundtruths: List[Tensor]
     groundtruth_labels: List[Tensor]
-    groundtruth_masks: List[Tensor]
-    detection_masks: List[Tensor]
 
     def __init__(
         self,
@@ -303,13 +301,11 @@ def __init__(
             raise ValueError("Expected argument `class_metrics` to be a boolean")
 
         self.class_metrics = class_metrics
-        self.add_state("detection_boxes", default=[], dist_reduce_fx=None)
+        self.add_state("detections", default=[], dist_reduce_fx=None)
         self.add_state("detection_scores", default=[], dist_reduce_fx=None)
         self.add_state("detection_labels", default=[], dist_reduce_fx=None)
-        self.add_state("groundtruth_boxes", default=[], dist_reduce_fx=None)
+        self.add_state("groundtruths", default=[], dist_reduce_fx=None)
         self.add_state("groundtruth_labels", default=[], dist_reduce_fx=None)
-        self.add_state("detection_masks", default=[], dist_reduce_fx=None)
-        self.add_state("groundtruth_masks", default=[], dist_reduce_fx=None)
 
     def update(self, preds: List[Dict[str, Tensor]], target: List[Dict[str, Tensor]]) -> None:  # type: ignore
         """Add detections and ground truth to the metric.
@@ -354,32 +350,29 @@ def update(self, preds: List[Dict[str, Tensor]], target: List[Dict[str, Tensor]]
             ValueError:
                 If any score is not type float and of length 1
         """
-        _input_validator(preds, target)
+        _input_validator(preds, target, iou_type=self.iou_type)
 
         for item in preds:
-            boxes, masks = self._get_safe_item_values(item)
-            self.detection_boxes.append(boxes)
+            detections = self._get_safe_item_values(item)
+            self.detections.append(detections)
             self.detection_labels.append(item["labels"])
             self.detection_scores.append(item["scores"])
-            self.detection_masks.append(masks)
 
         for item in target:
-            boxes, masks = self._get_safe_item_values(item)
-            self.groundtruth_boxes.append(boxes)
+            groundtruths = self._get_safe_item_values(item)
+            self.groundtruths.append(groundtruths)
             self.groundtruth_labels.append(item["labels"])
-            self.groundtruth_masks.append(masks)
 
     def _get_safe_item_values(self, item):
         if self.iou_type == "bbox":
             boxes = _fix_empty_tensors(item["boxes"])
             boxes = box_convert(boxes, in_fmt=self.box_format, out_fmt="xyxy")
-            masks = _fix_empty_tensors(torch.Tensor())
-        elif self.iou_type == "masks":
+            return boxes
+        elif self.iou_type == "segm":
             masks = _fix_empty_tensors(item["masks"])
-            boxes = _fix_empty_tensors(torch.Tensor())
+            return masks
         else:
             raise Exception(f"IOU type {self.iou_type} is not supported")
-        return boxes, masks
 
     def _get_classes(self) -> List:
         """Returns a list of unique classes found in ground truth and detection data."""
@@ -388,18 +381,11 @@ def _get_classes(self) -> List:
         return []
 
     def _compute_iou(self, id: int, class_id: int, max_det: int) -> Tensor:
-        return self._compute_iou_impl(id, self.groundtruth_boxes, self.detection_boxes, class_id, max_det, box_iou)
+        iou_func = box_iou if self.iou_type == "bbox" else segm_iou
 
-        # if self.iou_type == "segm":
-        #     return self._compute_iou_impl(id, self.groundtruth_masks, self.detection_masks, class_id, max_det, segm_iou)
-        # elif self.iou_type == "bbox":
+        return self._compute_iou_impl(id, class_id, max_det, iou_func)
 
-        # else:
-        #     raise Exception(f"IOU type {self.iou_type} is not supported")
-
-    def _compute_iou_impl(
-        self, id: int, ground_truths, detections, class_id: int, max_det: int, compute_iou: Callable
-    ) -> Tensor:
+    def _compute_iou_impl(self, id: int, class_id: int, max_det: int, compute_iou: Callable) -> Tensor:
         """Computes the Intersection over Union (IoU) for ground truth and detection bounding boxes for the given
         image and class.
 
@@ -412,8 +398,8 @@ def _compute_iou_impl(
                 Maximum number of evaluated detection bounding boxes
         """
         # if self.iou_type == "bbox":
-        gt = self.groundtruth_boxes[id]
-        det = self.detection_boxes[id]
+        gt = self.groundtruths[id]
+        det = self.detections[id]
 
         gt_label_mask = self.groundtruth_labels[id] == class_id
         det_label_mask = self.detection_labels[id] == class_id
@@ -452,8 +438,8 @@ def _evaluate_image(
             ious:
                 IoU results for image and class.
         """
-        gt = self.groundtruth_boxes[id]
-        det = self.detection_boxes[id]
+        gt = self.groundtruths[id]
+        det = self.detections[id]
         gt_label_mask = self.groundtruth_labels[id] == class_id
         det_label_mask = self.detection_labels[id] == class_id
         if len(gt_label_mask) == 0 or len(det_label_mask) == 0:
@@ -595,7 +581,7 @@ def _calculate(self, class_ids: List) -> Tuple[MAPMetricResults, MARMetricResult
             class_ids:
                 List of label class Ids.
         """
-        img_ids = range(len(self.groundtruth_boxes))
+        img_ids = range(len(self.groundtruths))
         max_detections = self.max_detection_thresholds[-1]
         area_ranges = self.bbox_area_ranges.values()
 
@@ -766,13 +752,11 @@ def compute(self) -> dict:
         """
 
         # move everything to CPU, as we are faster here
-        self.detection_boxes = [box.cpu() for box in self.detection_boxes]
+        self.detections = [box.cpu() for box in self.detections]
         self.detection_labels = [label.cpu() for label in self.detection_labels]
         self.detection_scores = [score.cpu() for score in self.detection_scores]
-        self.groundtruth_boxes = [box.cpu() for box in self.groundtruth_boxes]
+        self.groundtruths = [box.cpu() for box in self.groundtruths]
         self.groundtruth_labels = [label.cpu() for label in self.groundtruth_labels]
-        self.groundtruth_masks = [box.cpu() for box in self.groundtruth_masks]
-        self.detection_masks = [label.cpu() for label in self.detection_masks]
 
         classes = self._get_classes()
         precisions, recalls = self._calculate(classes)