Include empty images in exported annotations (#1479)

CHANGELOG.md

@@ -11,7 +11,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 - Downloaded file name in annotations export became more informative (https://github.com/opencv/cvat/pull/1352)
-- Added auto trimming for trailing whitespaces style enforsement (https://github.com/opencv/cvat/pull/1352)
+- Added auto trimming for trailing whitespaces style enforcement (https://github.com/opencv/cvat/pull/1352)
 - REST API: updated `GET /task/<id>/annotations`: parameters are `format`, `filename` (now optional), `action` (optional) (https://github.com/opencv/cvat/pull/1352)
 - REST API: removed `dataset/formats`, changed format of `annotation/formats` (https://github.com/opencv/cvat/pull/1352)
 - Exported annotations are stored for N hours instead of indefinitely (https://github.com/opencv/cvat/pull/1352)
@@ -20,6 +20,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Formats: most of formats renamed, no extension in title (https://github.com/opencv/cvat/pull/1352)
 - Formats: definitions are changed, are not stored in DB anymore (https://github.com/opencv/cvat/pull/1352)
 - cvat-core: session.annotations.put() now returns identificators of added objects (https://github.com/opencv/cvat/pull/1493)
+- Images without annotations now also included in dataset/annotations export (https://github.com/opencv/cvat/issues/525)
 
 ### Deprecated
 -

cvat/apps/dataset_manager/bindings.py

@@ -413,7 +413,7 @@ class CvatTaskDataExtractor(datumaro.SourceExtractor):
         if include_images:
             frame_provider = FrameProvider(task_data.db_task.data)
 
-        for frame_data in task_data.group_by_frame(include_empty=include_images):
+        for frame_data in task_data.group_by_frame(include_empty=True):
             loader = None
             if include_images:
                 loader = lambda p, i=frame_data.idx: frame_provider.get_frame(i,

cvat/apps/dataset_manager/formats/cvat.py

@@ -175,7 +175,7 @@ def dump_as_cvat_annotation(file_object, annotations):
     dumper.open_root()
     dumper.add_meta(annotations.meta)
 
-    for frame_annotation in annotations.group_by_frame():
+    for frame_annotation in annotations.group_by_frame(include_empty=True):
         frame_id = frame_annotation.frame
         dumper.open_image(OrderedDict([
             ("id", str(frame_id)),

cvat/apps/dataset_manager/formats/mot.py

@@ -39,6 +39,7 @@ def _import(src_file, task_data):
         label_cat = dataset.categories()[datumaro.AnnotationType.label]
 
         for item in dataset:
+            item = item.wrap(id=int(item.id) - 1) # NOTE: MOT frames start from 1
             frame_id = match_frame(item, task_data)
 
             for ann in item.annotations:

cvat/apps/dataset_manager/tests/_test_formats.py

@@ -51,6 +51,9 @@ def _setUpModule():
     import cvat.apps.dataset_manager as dm
     globals()['dm'] = dm
 
+    import datumaro
+    globals()['datumaro'] = datumaro
+
     import sys
     sys.path.insert(0, __file__[:__file__.rfind('/dataset_manager/')])
 
@@ -61,6 +64,7 @@ from io import BytesIO
 import os.path as osp
 import random
 import tempfile
+import zipfile
 
 from PIL import Image
 from django.contrib.auth.models import User, Group
@@ -113,38 +117,7 @@ class TaskExportTest(APITestCase):
     def setUpTestData(cls):
         create_db_users(cls)
 
-    def _generate_task(self):
-        task = {
-            "name": "my task #1",
-            "owner": '',
-            "assignee": '',
-            "overlap": 0,
-            "segment_size": 100,
-            "z_order": False,
-            "labels": [
-                {
-                    "name": "car",
-                    "attributes": [
-                        {
-                            "name": "model",
-                            "mutable": False,
-                            "input_type": "select",
-                            "default_value": "mazda",
-                            "values": ["bmw", "mazda", "renault"]
-                        },
-                        {
-                            "name": "parked",
-                            "mutable": True,
-                            "input_type": "checkbox",
-                            "default_value": False
-                        },
-                    ]
-                },
-                {"name": "person"},
-            ]
-        }
-        task = self._create_task(task, 3)
-
+    def _generate_annotations(self, task):
         annotations = {
             "version": 0,
             "tags": [
@@ -256,8 +229,39 @@ class TaskExportTest(APITestCase):
             ]
         }
         self._put_api_v1_task_id_annotations(task["id"], annotations)
+        return annotations
 
-        return task, annotations
+    def _generate_task(self):
+        task = {
+            "name": "my task #1",
+            "owner": '',
+            "assignee": '',
+            "overlap": 0,
+            "segment_size": 100,
+            "z_order": False,
+            "labels": [
+                {
+                    "name": "car",
+                    "attributes": [
+                        {
+                            "name": "model",
+                            "mutable": False,
+                            "input_type": "select",
+                            "default_value": "mazda",
+                            "values": ["bmw", "mazda", "renault"]
+                        },
+                        {
+                            "name": "parked",
+                            "mutable": True,
+                            "input_type": "checkbox",
+                            "default_value": False
+                        },
+                    ]
+                },
+                {"name": "person"},
+            ]
+        }
+        return self._create_task(task, 3)
 
     def _create_task(self, data, size):
         with ForceLogin(self.user, self.client):
@@ -285,53 +289,99 @@ class TaskExportTest(APITestCase):
 
         return response
 
-    def _test_export(self, format_name, save_images=False):
-        task, _ = self._generate_task()
-
+    def _test_export(self, check, task, format_name, **export_args):
         with tempfile.TemporaryDirectory() as temp_dir:
             file_path = osp.join(temp_dir, format_name)
             dm.task.export_task(task["id"], file_path,
-                format_name, save_images=save_images)
-
-            with open(file_path, 'rb') as f:
-                self.assertTrue(len(f.read()) != 0)
-
-    def test_datumaro(self):
-        self._test_export('Datumaro 1.0', save_images=False)
-
-    def test_coco(self):
-        self._test_export('COCO 1.0', save_images=True)
-
-    def test_voc(self):
-        self._test_export('PASCAL VOC 1.1', save_images=True)
-
-    def test_tf_record(self):
-        self._test_export('TFRecord 1.0', save_images=True)
+                format_name, **export_args)
 
-    def test_yolo(self):
-        self._test_export('YOLO 1.1', save_images=True)
-
-    def test_mot(self):
-        self._test_export('MOT 1.1', save_images=True)
-
-    def test_labelme(self):
-        self._test_export('LabelMe 3.0', save_images=True)
-
-    def test_mask(self):
-        self._test_export('Segmentation mask 1.1', save_images=True)
-
-    def test_cvat_video(self):
-        self._test_export('CVAT for video 1.1', save_images=True)
-
-    def test_cvat_images(self):
-        self._test_export('CVAT for images 1.1', save_images=True)
+            check(file_path)
 
     def test_export_formats_query(self):
         formats = dm.views.get_export_formats()
 
-        self.assertEqual(len(formats), 10)
+        self.assertEqual({f.DISPLAY_NAME for f in formats},
+        {
+            'COCO 1.0',
+            'CVAT for images 1.1',
+            'CVAT for video 1.1',
+            'Datumaro 1.0',
+            'LabelMe 3.0',
+            'MOT 1.1',
+            'PASCAL VOC 1.1',
+            'Segmentation mask 1.1',
+            'TFRecord 1.0',
+            'YOLO 1.1',
+        })
 
     def test_import_formats_query(self):
         formats = dm.views.get_import_formats()
 
-        self.assertEqual(len(formats), 8)
+        self.assertEqual({f.DISPLAY_NAME for f in formats},
+        {
+            'COCO 1.0',
+            'CVAT 1.1',
+            'LabelMe 3.0',
+            'MOT 1.1',
+            'PASCAL VOC 1.1',
+            'Segmentation mask 1.1',
+            'TFRecord 1.0',
+            'YOLO 1.1',
+        })
+
+    def test_exports(self):
+        def check(file_path):
+            with open(file_path, 'rb') as f:
+                self.assertTrue(len(f.read()) != 0)
+
+        for f in dm.views.get_export_formats():
+            format_name = f.DISPLAY_NAME
+            for save_images in { True, False }:
+                with self.subTest(format=format_name, save_images=save_images):
+                    task = self._generate_task()
+                    self._generate_annotations(task)
+                    self._test_export(check, task,
+                        format_name, save_images=save_images)
+
+    def test_empty_images_are_exported(self):
+        dm_env = dm.formats.registry.dm_env
+
+        for format_name, importer_name in [
+            ('COCO 1.0', 'coco'),
+            ('CVAT for images 1.1', 'cvat'),
+            # ('CVAT for video 1.1', 'cvat'), # does not support
+            ('Datumaro 1.0', 'datumaro_project'),
+            ('LabelMe 3.0', 'label_me'),
+            # ('MOT 1.1', 'mot_seq'), # does not support
+            ('PASCAL VOC 1.1', 'voc'),
+            ('Segmentation mask 1.1', 'voc'),
+            ('TFRecord 1.0', 'tf_detection_api'),
+            ('YOLO 1.1', 'yolo'),
+        ]:
+            with self.subTest(format=format_name):
+                task = self._generate_task()
+
+                def check(file_path):
+                    def load_dataset(src):
+                        if importer_name == 'datumaro_project':
+                            project = datumaro.components.project. \
+                                Project.load(src)
+
+                            # NOTE: can't import cvat.utils.cli
+                            # for whatever reason, so remove the dependency
+                            project.config.remove('sources')
+
+                            return project.make_dataset()
+                        return dm_env.make_importer(importer_name)(src) \
+                            .make_dataset()
+
+                    if zipfile.is_zipfile(file_path):
+                        with tempfile.TemporaryDirectory() as tmp_dir:
+                            zipfile.ZipFile(file_path).extractall(tmp_dir)
+                            dataset = load_dataset(tmp_dir)
+                    else:
+                        dataset = load_dataset(file_path)
+
+                    self.assertEqual(len(dataset), task["size"])
+                self._test_export(check, task, format_name, save_images=False)
+

datumaro/datumaro/components/project.py

@@ -142,7 +142,7 @@ def load_project_as_dataset(url):
 
 class Environment:
     _builtin_plugins = None
-    PROJECT_EXTRACTOR_NAME = 'project'
+    PROJECT_EXTRACTOR_NAME = 'datumaro_project'
 
     def __init__(self, config=None):
         config = Config(config,

datumaro/datumaro/plugins/yolo_format/extractor.py

@@ -93,12 +93,6 @@ class YoloExtractor(SourceExtractor):
                     (osp.splitext(osp.basename(p.strip()))[0], p.strip())
                     for p in f
                 )
-
-            for item_id, image_path in subset.items.items():
-                image_path = self._make_local_path(image_path)
-                if not osp.isfile(image_path) and item_id not in image_info:
-                    raise Exception("Can't find image '%s'" % item_id)
-
             subsets[subset_name] = subset
 
         self._subsets = subsets
@@ -122,10 +116,9 @@ class YoloExtractor(SourceExtractor):
             image_path = self._make_local_path(item)
             image_size = self._image_info.get(item_id)
             image = Image(path=image_path, size=image_size)
-            h, w = image.size
 
             anno_path = osp.splitext(image_path)[0] + '.txt'
-            annotations = self._parse_annotations(anno_path, w, h)
+            annotations = self._parse_annotations(anno_path, image)
 
             item = DatasetItem(id=item_id, subset=subset_name,
                 image=image, annotations=annotations)
@@ -134,21 +127,30 @@ class YoloExtractor(SourceExtractor):
         return item
 
     @staticmethod
-    def _parse_annotations(anno_path, image_width, image_height):
+    def _parse_annotations(anno_path, image):
+        lines = []
         with open(anno_path, 'r') as f:
-            annotations = []
             for line in f:
-                label_id, xc, yc, w, h = line.strip().split()
-                label_id = int(label_id)
-                w = float(w)
-                h = float(h)
-                x = float(xc) - w * 0.5
-                y = float(yc) - h * 0.5
-                annotations.append(Bbox(
-                    round(x * image_width, 1), round(y * image_height, 1),
-                    round(w * image_width, 1), round(h * image_height, 1),
-                    label=label_id
-                ))
+                line = line.strip()
+                if line:
+                    lines.append(line)
+
+        annotations = []
+        if lines:
+            image_height, image_width = image.size # use image info late
+        for line in lines:
+            label_id, xc, yc, w, h = line.split()
+            label_id = int(label_id)
+            w = float(w)
+            h = float(h)
+            x = float(xc) - w * 0.5
+            y = float(yc) - h * 0.5
+            annotations.append(Bbox(
+                round(x * image_width, 1), round(y * image_height, 1),
+                round(w * image_width, 1), round(h * image_height, 1),
+                label=label_id
+            ))
+
         return annotations
 
     @staticmethod