comment annotation

Author: hovnatan

docs/source/tutorial.sdk.rst

@@ -359,7 +359,7 @@ SuperAnnotate format annotations:
    df = sa.aggregate_annotations_as_df("<path_to_project_folder>")
 
 The created DataFrame will have columns :code:`imageName`, :code:`instanceId`,
-:code:`className`, :code:`attributeGroupName`, :code:`attributeName`, :code:`type`, :code:`error`, :code:`locked`, :code:`visible`, :code:`trackingId`, :code:`probability`, :code:`pointLabels`, :code:`meta` (geometry information as string), :code:`commentResolved`, :code:`classColor`.
+:code:`className`, :code:`attributeGroupName`, :code:`attributeName`, :code:`type`, :code:`error`, :code:`locked`, :code:`visible`, :code:`trackingId`, :code:`probability`, :code:`pointLabels`, :code:`meta` (geometry information as string), :code:`commentResolved`, :code:`classColor`, :code:`groupId`.
 
 Example of created DataFrame:
 

superannotate/__init__.py

@@ -21,6 +21,7 @@
     image_path_to_annotation_paths, project_type_int_to_str,
     project_type_str_to_int, user_role_str_to_int
 )
+from .dataframe_filtering import filter_annotation_instances, filter_comments
 from .db.annotation_classes import (
     create_annotation_class, create_annotation_classes_from_classes_json,
     delete_annotation_class, download_annotation_classes_json,
@@ -62,7 +63,6 @@
     convert_platform, convert_project_type, export_annotation_format,
     import_annotation_format
 )
-from .instance_filtering import filter_annotation_instances
 from .version import __version__
 
 formatter = logging.Formatter(fmt='SA-PYTHON-SDK - %(levelname)s - %(message)s')

superannotate/analytics/common.py

@@ -9,17 +9,15 @@
 
 def df_to_annotations(df, output_dir):
     """Converts and saves pandas DataFrame annotation info (see aggregate_annotations_as_df) in output_dir
-    The DataFrame should have columns: "imageName", "classNmae", "attributeGroupName", "attributeName", "type", "error", "locked", "visible", trackingId", "probability", "pointLabels", "meta", "commentResolved", "classColor"
+    The DataFrame should have columns: "imageName", "classNmae", "attributeGroupName", "attributeName", "type", "error", "locked", "visible", trackingId", "probability", "pointLabels", "meta", "commentResolved", "classColor", "groupId"
 
     Currently only works for Vector projects.
 
-    :param df: pandas DataFrame
+    :param df: pandas DataFrame of annotations possibly created by aggregate_annotations_as_df
     :type df: pandas.DataFrame
-    :param include_classes_wo_annotations: enables inclusion of classes without annotations info
-    :type include_classes_wo_annotations: bool
+    :param output_dir: output dir for annotations and classes.json
+    :type output_dir: str or Pathlike
 
-    :return: DataFrame on annotations with columns: ["imageName", "classNmae", "attributeGroupName", "attributeName", "type", "error", "locked", "visible", trackingId", "probability", "pointLabels", "meta", "commentResolved"]
-    :rtype: pandas DataFrame
     """
 
     project_suffix = "objects.json"
@@ -50,6 +48,7 @@ def df_to_annotations(df, output_dir):
             )
             instance_annotation["trackingId"] = instance_df.iloc[0]["trackingId"
                                                                    ]
+            instance_annotation["groupId"] = int(instance_df.iloc[0]["groupId"])
             instance_annotation.update(annotation_meta)
             for _, row in instance_df.iterrows():
                 if row["attributeGroupName"] is not None:
@@ -60,6 +59,14 @@ def df_to_annotations(df, output_dir):
                         }
                     )
             image_annotation.append(instance_annotation)
+
+        comments = image_df[image_df["type"] == "comment"]
+        for _, comment in comments.iterrows():
+            comment_json = {"type": "comment"}
+            comment_json.update(comment["meta"])
+            comment_json["resolved"] = comment["commentResolved"]
+            image_annotation.append(comment_json)
+
         json.dump(
             image_annotation,
             open(output_dir / f"{image}___{project_suffix}", "w"),
@@ -124,7 +131,7 @@ def aggregate_annotations_as_df(
     :param include_comments: enables inclusion of comments info as commentResolved column
     :type include_comments: bool
 
-    :return: DataFrame on annotations with columns: "imageName", "instanceId" className", "attributeGroupName", "attributeName", "type", "error", "locked", "visible", "trackingId", "probability", "pointLabels", "meta" (geometry information as string), "commentResolved", "classColor"
+    :return: DataFrame on annotations with columns: "imageName", "instanceId" className", "attributeGroupName", "attributeName", "type", "error", "locked", "visible", "trackingId", "probability", "pointLabels", "meta" (geometry information as string), "commentResolved", "classColor", "groupId"
     :rtype: pandas DataFrame
     """
 
@@ -148,6 +155,7 @@ def aggregate_annotations_as_df(
         "pointLabels": [],
         "meta": [],
         "classColor": [],
+        "groupId": []
     }
 
     if include_comments:
@@ -192,12 +200,9 @@ def __append_annotation(annotation_dict):
         annotation_image_name = annotation_path.name.split("___")[0]
         annotation_instance_id = 0
         for annotation in annotation_json:
-
             annotation_type = annotation.get("type", "mask")
-
             if annotation_type in ['meta', 'tag']:
                 continue
-
             if annotation_type == "comment":
                 if include_comments:
                     comment_resolved = annotation["resolved"]
@@ -215,26 +220,18 @@ def __append_annotation(annotation_dict):
                         }
                     )
                 continue
-
             annotation_instance_id += 1
-
             annotation_class_name = annotation.get("className")
-            
-            if annotation_class_name:
-                annotation_class_color = class_name_to_color[annotation_class_name]
-            else:
+            if annotation_class_name is None:
                 raise SABaseException(
                     0, "Annotation class not found in classes.json"
                 )
-
+            annotation_class_color = class_name_to_color[annotation_class_name]
+            annotation_group_id = annotation.get("groupId")
             annotation_locked = annotation.get("locked")
-
             annotation_visible = annotation.get("visible")
-
             annotation_tracking_id = annotation.get("trackingId")
-
             annotation_meta = None
-
             if annotation_type in ["bbox", "polygon", "polyline", "cuboid"]:
                 annotation_meta = {"points": annotation["points"]}
             elif annotation_type == "point":
@@ -254,13 +251,9 @@ def __append_annotation(annotation_dict):
                     "connections": annotation["connections"],
                     "points": annotation["points"]
                 }
-
             annotation_error = annotation.get('error')
-
             annotation_probability = annotation.get("probability")
-
             annotation_point_labels = annotation.get("pointLabels")
-
             attributes = annotation.get("attributes")
 
             if not attributes:
@@ -277,7 +270,8 @@ def __append_annotation(annotation_dict):
                         "error": annotation_error,
                         "probability": annotation_probability,
                         "pointLabels": annotation_point_labels,
-                        "classColor": annotation_class_color
+                        "classColor": annotation_class_color,
+                        "groupId": annotation_group_id
                     }
                 )
 
@@ -301,7 +295,8 @@ def __append_annotation(annotation_dict):
                         "error": annotation_error,
                         "probability": annotation_probability,
                         "pointLabels": annotation_point_labels,
-                        "classColor": annotation_class_color
+                        "classColor": annotation_class_color,
+                        "groupId": annotation_group_id
                     }
                 )
 

superannotate/instance_filtering.py superannotate/dataframe_filtering.pysuperannotate/instance_filtering.py renamed to superannotate/dataframe_filtering.py

@@ -1,6 +1,24 @@
 import pandas as pd
 
 
+def filter_comments(annotations_df, resolved=False):
+    """Filter comments on resolve status
+
+    :param annotations_df: pandas DataFrame of project annotations
+    :type annotations_df: pandas.DataFrame
+    :param resolved: resolved criterion
+    :type resolved: bool
+
+    :return: filtered DataFrame
+    :rtype: pandas.DataFrame
+
+    """
+    df = annotations_df[annotations_df["type"] == "comment"]
+    df = df[df["commentResolved"] == resolved]
+
+    return df
+
+
 def filter_annotation_instances(annotations_df, include=None, exclude=None):
     """Filter annotation instances from project annotations pandas DataFrame.
 

superannotate/version.py

@@ -1 +1 @@
-__version__ = "2.1.2"
+__version__ = "2.1.3"

tests/sample_project_vector/example_image_1.jpg___objects.json

@@ -1,4 +1,16 @@
 [
+    {
+        "type": "comment",
+        "x": 621.4069213867188,
+        "y": 631.5987548828125,
+        "comments": [
+            {
+                "text": "Bordyuri mi mas@ petqa lini parking class-i mej myus mas@ Terrian class-i",
+                "id": "hovnatan@superannotate.com"
+            }
+        ],
+        "resolved": true
+    },
     {
         "type": "bbox",
         "classId": 55917,

tests/test_annotation_adding.py

@@ -87,11 +87,7 @@ def test_add_bbox(tmpdir):
     export = sa.prepare_export(project, include_fuse=True)
     sa.download_export(project, export, tmpdir)
 
-    annotations_new_export = json.load(
-        open(tmpdir / f"{image_name}___objects.json")
-    )
-
-    df = sa.aggregate_annotations_as_df(tmpdir)
+    df = sa.aggregate_annotations_as_df(tmpdir, include_comments=True)
 
     num = len(df[df["imageName"] == image_name]["instanceId"].unique())
 

tests/test_basic_images.py

@@ -71,12 +71,13 @@ def test_basic_images(project_type, name, description, from_folder, tmpdir):
     downloaded_classes = json.load(open(tmpdir / "classes.json"))
 
     for a in annotation:
-        found = False
+        if "className" not in a:
+            continue
         for c1 in downloaded_classes:
             if a["className"] == c1["name"]:
-                found = True
                 break
-        assert found
+        else:
+            assert False
 
     input_classes = json.load(open(from_folder / "classes" / "classes.json"))
     assert len(downloaded_classes) == len(input_classes)

tests/test_filter_instances.py

@@ -7,6 +7,23 @@
 PROJECT_DIR = "./tests/sample_project_vector"
 
 
+def test_filter_comments(tmpdir):
+    tmpdir = Path(tmpdir)
+
+    # projects = sa.search_projects(PROJECT_NAME_1, return_metadata=True)
+    # for project in projects:
+    #     sa.delete_project(project)
+
+    # project = sa.create_project(PROJECT_NAME_1, "test", "Vector")
+
+    not_filtered = sa.aggregate_annotations_as_df(
+        PROJECT_DIR, include_comments=True
+    )
+
+    filtered_excl = sa.filter_comments(not_filtered, True)
+    print(str(filtered_excl["meta"].iloc[0]))
+
+
 def test_filter_instances(tmpdir):
     tmpdir = Path(tmpdir)
 
@@ -89,12 +106,11 @@ def test_df_to_annotations(tmpdir):
     # print(df_new["image_name"].value_counts())
     # print(df["image_name"].value_counts())
     for _index, row in enumerate(df.iterrows()):
-        found = False
         for _, row_2 in enumerate(df_new.iterrows()):
             if row_2[1].equals(row[1]):
-                found = True
                 break
-        assert found
+        else:
+            assert False
     for project in sa.search_projects("test df to annotations 2"):
         sa.delete_project(project)
     project = sa.create_project("test df to annotations 2", "test", "Vector")
@@ -113,12 +129,20 @@ def test_df_to_annotations_full(tmpdir):
     tmpdir = Path(tmpdir)
 
     df = sa.aggregate_annotations_as_df(
-        PROJECT_DIR, include_classes_wo_annotations=True
+        PROJECT_DIR, include_classes_wo_annotations=True, include_comments=True
     )
     sa.df_to_annotations(df, tmpdir)
     df_new = sa.aggregate_annotations_as_df(
-        tmpdir, include_classes_wo_annotations=True
+        tmpdir, include_classes_wo_annotations=True, include_comments=True
+    )
+    for project in sa.search_projects("test df to annotations 4"):
+        sa.delete_project(project)
+    project = sa.create_project("test df to annotations 4", "test", "Vector")
+    sa.upload_images_from_folder_to_project(project, PROJECT_DIR)
+    sa.create_annotation_classes_from_classes_json(
+        project, tmpdir / "classes" / "classes.json"
     )
+    sa.upload_annotations_from_folder_to_project(project, tmpdir)
     # print(df_new["image_name"].value_counts())
     # print(df["image_name"].value_counts())
     for _index, row in enumerate(df.iterrows()):