Forward CVS-94422 size bug fix PR to release branch

data/car_tree_bug/annotations/hierarchical_default.json

@@ -1,23 +1,24 @@
-{"images": [["Slide18.PNG", ["car"]], 
-            ["Slide19.PNG", ["tree"]], 
-            ["Slide16.PNG", ["car", "tree"]],
-            ["Slide20.PNG", ["bug"]], 
-            ["Slide2.PNG", ["car", "tree"]], 
-            ["Slide15.PNG", ["bug", "car"]]
-            ],
- "hierarchy":
-            [
-                {
-                "parent": "self",
-                "group": "natural",
-                "labels": ["tree", "bug"],
-                "task_type": "single-label"
-                },
-                {
-                "parent": "self",
-                "group": "manmade",
-                "labels": ["car"],
-                "task_type": "multi-label"
-                }
-            ]
+{
+  "images": [
+    ["Slide18.PNG", ["car"]],
+    ["Slide19.PNG", ["tree"]],
+    ["Slide16.PNG", ["car", "tree"]],
+    ["Slide20.PNG", ["bug"]],
+    ["Slide2.PNG", ["car", "tree"]],
+    ["Slide15.PNG", ["bug", "car"]]
+  ],
+  "hierarchy": [
+    {
+      "parent": "self",
+      "group": "natural",
+      "labels": ["tree", "bug"],
+      "task_type": "single-label"
+    },
+    {
+      "parent": "self",
+      "group": "manmade",
+      "labels": ["car"],
+      "task_type": "multi-label"
+    }
+  ]
 }

external/model-preparation-algorithm/mpa_tasks/apis/detection/task.py

@@ -340,12 +340,13 @@ def _det_add_predictions_to_dataset(self, all_results, width, height, confidence
             for i in range(detections.shape[0]):
                 probability = float(detections[i, 4])
                 coords = detections[i, :4].astype(float).copy()
-                coords /= np.array([width, height, width, height], dtype=float)
-                coords = np.clip(coords, 0, 1)
 
-                if probability < confidence_threshold:
+                if probability < confidence_threshold or (coords[2] - coords[0]) * (coords[3] - coords[1]) < 1.0:
                     continue
 
+                coords /= np.array([width, height, width, height], dtype=float)
+                coords = np.clip(coords, 0, 1)
+
                 assigned_label = [ScoredLabel(self._labels[label_idx],
                                               probability=probability)]
                 if coords[3] - coords[1] <= 0 or coords[2] - coords[0] <= 0:
@@ -368,7 +369,7 @@ def _ins_seg_add_predictions_to_dataset(self, all_results, width, height, confid
                 for contour, hierarchy in zip(contours, hierarchies[0]):
                     if hierarchy[3] != -1:
                         continue
-                    if len(contour) <= 2 or probability < confidence_threshold:
+                    if len(contour) <= 2 or probability < confidence_threshold or cv2.contourArea(contour) < 1.0:
                         continue
                     if self._task_type == TaskType.INSTANCE_SEGMENTATION:
                         points = [Point(x=point[0][0] / width, y=point[0][1] / height) for point in contour]
@@ -377,8 +378,7 @@ def _ins_seg_add_predictions_to_dataset(self, all_results, width, height, confid
                         points = [Point(x=point[0] / width, y=point[1] / height) for point in box_points]
                     labels = [ScoredLabel(self._labels[label_idx], probability=probability)]
                     polygon = Polygon(points=points)
-                    if cv2.contourArea(contour) > 0 and polygon.get_area() > 1e-12:
-                        shapes.append(Annotation(polygon, labels=labels, id=ID(f"{label_idx:08}")))
+                    shapes.append(Annotation(polygon, labels=labels, id=ID(f"{label_idx:08}")))
         return shapes
 
     @staticmethod

ote_cli/ote_cli/utils/hpo.py

@@ -237,9 +237,9 @@ def run_hpo_trainer(
     # need to set batch_size config from the instance of ConfigurableParameters
     if hp_config["batch_size_param_name"] not in hp_config["params"].keys():
         attr = hyper_parameters
-        for val in hp_config["batch_size_param_name"].split('.'):
+        for val in hp_config["batch_size_param_name"].split("."):
             attr = getattr(attr, val)
-        hp_config["batch_size"] =  attr
+        hp_config["batch_size"] = attr
 
     # set hyper-parameters and print them
     HpoManager.set_hyperparameter(hyper_parameters, hp_config["params"])

ote_sdk/ote_sdk/tests/usecases/exportable_code/test_visualization.py

@@ -107,7 +107,7 @@ def check_visualizer_attributes(
             expected_delay: int,
             expected_show_count: bool,
             expected_is_one_label: bool,
-            expected_no_show: bool
+            expected_no_show: bool,
         ):
             assert actual_visualizer.window_name == expected_name
             assert actual_visualizer.delay == expected_delay
@@ -169,7 +169,6 @@ def test_visualizer_draw(self):
         actual_image = Visualizer().draw(image=image, annotation=annotation_scene)
         assert np.array_equal(actual_image, expected_image)
 
-
     @pytest.mark.priority_medium
     @pytest.mark.unit
     @pytest.mark.reqids(Requirements.REQ_1)
@@ -189,4 +188,3 @@ def test_visualizer_no_show_mode(self):
         visualizer = Visualizer(no_show=True)
         visualizer.show(image)
         visualizer.is_quit()
-

ote_sdk/ote_sdk/usecases/exportable_code/prediction_to_annotation_converter.py

@@ -177,7 +177,10 @@ def convert_to_annotation(
         image_size = metadata["original_shape"][1::-1]
         for box in predictions:
             scored_label = ScoredLabel(self.labels[int(box.id)], float(box.score))
-            coords = np.array(box.get_coords(), dtype=float) / np.tile(image_size, 2)
+            coords = np.array(box.get_coords(), dtype=float)
+            if (coords[2] - coords[0]) * (coords[3] - coords[1]) < 1.0:
+                continue
+            coords /= np.tile(image_size, 2)
             annotations.append(
                 Annotation(
                     Rectangle(coords[0], coords[1], coords[2], coords[3]),
@@ -376,9 +379,9 @@ def convert_to_annotation(
             for contour, hierarchy in zip(contours, hierarchies[0]):
                 if hierarchy[3] != -1:
                     continue
-                contour = list(contour)
-                if len(contour) <= 2:
+                if len(contour) <= 2 or cv2.contourArea(contour):
                     continue
+                contour = list(contour)
                 points = [
                     Point(
                         x=point[0][0] / metadata["original_shape"][1],
@@ -387,17 +390,14 @@ def convert_to_annotation(
                     for point in contour
                 ]
                 polygon = Polygon(points=points)
-                if polygon.get_area() > 1e-12:
-                    annotations.append(
-                        Annotation(
-                            polygon,
-                            labels=[
-                                ScoredLabel(
-                                    self.labels[int(class_idx) - 1], float(score)
-                                )
-                            ],
-                        )
+                annotations.append(
+                    Annotation(
+                        polygon,
+                        labels=[
+                            ScoredLabel(self.labels[int(class_idx) - 1], float(score))
+                        ],
                     )
+                )
         annotation_scene = AnnotationSceneEntity(
             kind=AnnotationSceneKind.PREDICTION,
             annotations=annotations,
@@ -427,7 +427,7 @@ def convert_to_annotation(
             for contour, hierarchy in zip(contours, hierarchies[0]):
                 if hierarchy[3] != -1:
                     continue
-                if len(contour) <= 2:
+                if len(contour) <= 2 or cv2.contourArea(contour):
                     continue
                 points = [
                     Point(
@@ -437,17 +437,14 @@ def convert_to_annotation(
                     for point in cv2.boxPoints(cv2.minAreaRect(contour))
                 ]
                 polygon = Polygon(points=points)
-                if polygon.get_area() > 1e-12:
-                    annotations.append(
-                        Annotation(
-                            polygon,
-                            labels=[
-                                ScoredLabel(
-                                    self.labels[int(class_idx) - 1], float(score)
-                                )
-                            ],
-                        )
+                annotations.append(
+                    Annotation(
+                        polygon,
+                        labels=[
+                            ScoredLabel(self.labels[int(class_idx) - 1], float(score))
+                        ],
                     )
+                )
         annotation_scene = AnnotationSceneEntity(
             kind=AnnotationSceneKind.PREDICTION,
             annotations=annotations,

ote_sdk/ote_sdk/usecases/exportable_code/visualizers/anomaly_visualizer.py

@@ -43,8 +43,9 @@ def __init__(
                 cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO | cv2.WINDOW_GUI_EXPANDED,
             )
             self.trackbar_name = "Opacity"
-            cv2.createTrackbar(self.trackbar_name, self.window_name,
-                               0, 100, lambda x: x)
+            cv2.createTrackbar(
+                self.trackbar_name, self.window_name, 0, 100, lambda x: x
+            )
 
     @staticmethod
     def to_heat_mask(mask: np.ndarray) -> np.ndarray:

ote_sdk/ote_sdk/utils/vis_utils.py

@@ -1,14 +1,17 @@
 # Copyright (C) 2021-2022 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
-from typing import Union, Iterable
+from typing import Iterable, Union
 
 import cv2
 import numpy as np
 
 
-def get_actmap(saliency_map: Union[np.ndarray, Iterable, int, float], output_res: Union[tuple, list]) -> np.ndarray:
-    """ Get activation map (heatmap)  from saliency map
+def get_actmap(
+    saliency_map: Union[np.ndarray, Iterable, int, float],
+    output_res: Union[tuple, list],
+) -> np.ndarray:
+    """Get activation map (heatmap)  from saliency map
 
     Args:
         saliency_map (Union[np.ndarray, Iterable, int, float]): Saliency map with pixel values from 0-255