Fix export job dataset (#5052)

CHANGELOG.md

@@ -47,6 +47,7 @@ non-ascii paths while adding files from "Connected file share" (issue #4428)
   (<https://github.com/opencv/cvat/pull/5057>)
 - Double modal export/backup a task/project (<https://github.com/opencv/cvat/pull/5075>)
 - Fixed bug of computing Job's unsolved/resolved issues numbers (<https://github.com/opencv/cvat/pull/5101>)
+- Dataset export for job (<https://github.com/opencv/cvat/pull/5052>)
 
 ### Security
 - TDB

cvat/apps/dataset_manager/formats/cvat.py

@@ -21,7 +21,7 @@ from datumaro.components.extractor import (DEFAULT_SUBSET_NAME, Extractor,
 from datumaro.util.image import Image
 from defusedxml import ElementTree
 
-from cvat.apps.dataset_manager.bindings import (ProjectData, TaskData,
+from cvat.apps.dataset_manager.bindings import (ProjectData, CommonData,
                                                 get_defaulted_subset,
                                                 import_dm_annotations,
                                                 match_dm_item)
@@ -984,11 +984,11 @@ def dump_as_cvat_interpolation(dumper, annotations):
         counter += 1
 
     for shape in annotations.shapes:
-        frame_step = annotations.frame_step if isinstance(annotations, TaskData) else annotations.frame_step[shape.task_id]
-        if isinstance(annotations, TaskData):
-            stop_frame = int(annotations.meta['task']['stop_frame'])
+        frame_step = annotations.frame_step if not isinstance(annotations, ProjectData) else annotations.frame_step[shape.task_id]
+        if not isinstance(annotations, ProjectData):
+            stop_frame = int(annotations.meta[annotations.META_FIELD]['stop_frame'])
         else:
-            task_meta = list(filter(lambda task: int(task[1]['id']) == shape.task_id, annotations.meta['project']['tasks']))[0][1]
+            task_meta = list(filter(lambda task: int(task[1]['id']) == shape.task_id, annotations.meta[annotations.META_FIELD]['tasks']))[0][1]
             stop_frame = int(task_meta['stop_frame'])
         track = {
             'label': shape.label,
@@ -1102,7 +1102,7 @@ def load_anno(file_object, annotations):
                         attributes={'frame': el.attrib['id']},
                         image=el.attrib['name']
                     ),
-                    task_data=annotations
+                    instance_data=annotations
                 ))
             elif el.tag in supported_shapes and (track is not None or image_is_opened):
                 if shape and shape['type'] == 'skeleton':
@@ -1258,10 +1258,10 @@ def load_anno(file_object, annotations):
                 tag = None
             el.clear()
 
-def dump_task_anno(dst_file, task_data, callback):
+def dump_task_or_job_anno(dst_file, instance_data, callback):
     dumper = create_xml_dumper(dst_file)
     dumper.open_document()
-    callback(dumper, task_data)
+    callback(dumper, instance_data)
     dumper.close_document()
 
 def dump_project_anno(dst_file: BufferedWriter, project_data: ProjectData, callback: Callable):
@@ -1270,33 +1270,34 @@ def dump_project_anno(dst_file: BufferedWriter, project_data: ProjectData, callb
     callback(dumper, project_data)
     dumper.close_document()
 
-def dump_media_files(task_data: TaskData, img_dir: str, project_data: ProjectData = None):
+def dump_media_files(instance_data: CommonData, img_dir: str, project_data: ProjectData = None):
     ext = ''
-    if task_data.meta['task']['mode'] == 'interpolation':
+    if instance_data.meta[instance_data.META_FIELD]['mode'] == 'interpolation':
         ext = FrameProvider.VIDEO_FRAME_EXT
 
-    frame_provider = FrameProvider(task_data.db_task.data)
+    frame_provider = FrameProvider(instance_data.db_data)
     frames = frame_provider.get_frames(
+        instance_data.start, instance_data.stop,
         frame_provider.Quality.ORIGINAL,
         frame_provider.Type.BUFFER)
-    for frame_id, (frame_data, _) in enumerate(frames):
-        if (project_data is not None and (task_data.db_task.id, frame_id) in project_data.deleted_frames) \
-            or frame_id in task_data.deleted_frames:
+    for frame_id, (frame_data, _) in zip(instance_data.rel_range, frames):
+        if (project_data is not None and (instance_data.db_instance.id, frame_id) in project_data.deleted_frames) \
+            or frame_id in instance_data.deleted_frames:
             continue
-        frame_name = task_data.frame_info[frame_id]['path'] if project_data is None \
-            else project_data.frame_info[(task_data.db_task.id, frame_id)]['path']
+        frame_name = instance_data.frame_info[frame_id]['path'] if project_data is None \
+            else project_data.frame_info[(instance_data.db_instance.id, frame_id)]['path']
         img_path = osp.join(img_dir, frame_name + ext)
         os.makedirs(osp.dirname(img_path), exist_ok=True)
         with open(img_path, 'wb') as f:
             f.write(frame_data.getvalue())
 
-def _export_task(dst_file, task_data, anno_callback, save_images=False):
+def _export_task_or_job(dst_file, instance_data, anno_callback, save_images=False):
     with TemporaryDirectory() as temp_dir:
         with open(osp.join(temp_dir, 'annotations.xml'), 'wb') as f:
-            dump_task_anno(f, task_data, anno_callback)
+            dump_task_or_job_anno(f, instance_data, anno_callback)
 
         if save_images:
-            dump_media_files(task_data, osp.join(temp_dir, 'images'))
+            dump_media_files(instance_data, osp.join(temp_dir, 'images'))
 
         make_zip_archive(temp_dir, dst_file)
 
@@ -1307,7 +1308,7 @@ def _export_project(dst_file: str, project_data: ProjectData, anno_callback: Cal
 
         if save_images:
             for task_data in project_data.task_data:
-                subset = get_defaulted_subset(task_data.db_task.subset, project_data.subsets)
+                subset = get_defaulted_subset(task_data.db_instance.subset, project_data.subsets)
                 subset_dir = osp.join(temp_dir, 'images', subset)
                 os.makedirs(subset_dir, exist_ok=True)
                 dump_media_files(task_data, subset_dir, project_data)
@@ -1320,7 +1321,7 @@ def _export_video(dst_file, instance_data, save_images=False):
         _export_project(dst_file, instance_data,
             anno_callback=dump_as_cvat_interpolation, save_images=save_images)
     else:
-        _export_task(dst_file, instance_data,
+        _export_task_or_job(dst_file, instance_data,
             anno_callback=dump_as_cvat_interpolation, save_images=save_images)
 
 @exporter(name='CVAT for images', ext='ZIP', version='1.1')
@@ -1329,7 +1330,7 @@ def _export_images(dst_file, instance_data, save_images=False):
         _export_project(dst_file, instance_data,
             anno_callback=dump_as_cvat_annotation, save_images=save_images)
     else:
-        _export_task(dst_file, instance_data,
+        _export_task_or_job(dst_file, instance_data,
             anno_callback=dump_as_cvat_annotation, save_images=save_images)
 
 @importer(name='CVAT', ext='XML, ZIP', version='1.1')

cvat/apps/dataset_manager/formats/kitti.py

 # Copyright (C) 2021-2022 Intel Corporation
+# Copyright (C) 2022 CVAT.ai Corporation
 #
 # SPDX-License-Identifier: MIT
 
@@ -9,8 +10,7 @@ from datumaro.components.dataset import Dataset
 from datumaro.plugins.kitti_format.format import KittiPath, write_label_map
 from pyunpack import Archive
 
-from cvat.apps.dataset_manager.bindings import (GetCVATDataExtractor,
-    ProjectData, import_dm_annotations)
+from cvat.apps.dataset_manager.bindings import (GetCVATDataExtractor, import_dm_annotations)
 from cvat.apps.dataset_manager.util import make_zip_archive
 
 from .transformations import RotatedBoxesToPolygons
@@ -45,8 +45,7 @@ def _import(src_file, instance_data):
             write_label_map(color_map_path, color_map)
 
         dataset = Dataset.import_from(tmp_dir, format='kitti', env=dm_env)
-        labels_meta = instance_data.meta['project']['labels'] \
-            if isinstance(instance_data, ProjectData) else instance_data.meta['task']['labels']
+        labels_meta = instance_data.meta[instance_data.META_FIELD]['labels']
         if 'background' not in [label['name'] for _, label in labels_meta]:
             dataset.filter('/item/annotation[label != "background"]',
                 filter_annotations=True)

cvat/apps/dataset_manager/formats/mot.py

 # Copyright (C) 2019-2022 Intel Corporation
+# Copyright (C) 2022 CVAT.ai Corporation
 #
 # SPDX-License-Identifier: MIT
 
@@ -13,13 +14,15 @@ from cvat.apps.dataset_manager.util import make_zip_archive
 
 from .registry import dm_env, exporter, importer
 
-def _import_task(dataset, task_data):
+def _import_to_task(dataset, instance_data):
     tracks = {}
     label_cat = dataset.categories()[datumaro.AnnotationType.label]
 
     for item in dataset:
-        frame_number = int(item.id) - 1 # NOTE: MOT frames start from 1
-        frame_number = task_data.abs_frame_id(frame_number)
+        # NOTE: MOT frames start from 1
+        # job has an offset, for task offset is 0
+        frame_number = int(item.id) - 1 + instance_data.start
+        frame_number = instance_data.abs_frame_id(frame_number)
 
         for ann in item.annotations:
             if ann.type != datumaro.AnnotationType.bbox:
@@ -28,7 +31,7 @@ def _import_task(dataset, task_data):
             track_id = ann.attributes.get('track_id')
             if track_id is None:
                 # Extension. Import regular boxes:
-                task_data.add_shape(task_data.LabeledShape(
+                instance_data.add_shape(instance_data.LabeledShape(
                     type='rectangle',
                     label=label_cat.items[ann.label].name,
                     points=ann.points,
@@ -41,7 +44,7 @@ def _import_task(dataset, task_data):
                 ))
                 continue
 
-            shape = task_data.TrackedShape(
+            shape = instance_data.TrackedShape(
                 type='rectangle',
                 points=ann.points,
                 occluded=ann.attributes.get('occluded') is True,
@@ -55,7 +58,7 @@ def _import_task(dataset, task_data):
 
             # build trajectories as lists of shapes in track dict
             if track_id not in tracks:
-                tracks[track_id] = task_data.Track(
+                tracks[track_id] = instance_data.Track(
                     label_cat.items[ann.label].name, 0, 'manual', [])
             tracks[track_id].shapes.append(shape)
 
@@ -67,10 +70,10 @@ def _import_task(dataset, task_data):
         prev_shape_idx = 0
         prev_shape = track.shapes[0]
         for shape in track.shapes[1:]:
-            has_skip = task_data.frame_step < shape.frame - prev_shape.frame
+            has_skip = instance_data.frame_step < shape.frame - prev_shape.frame
             if has_skip and not prev_shape.outside:
                 prev_shape = prev_shape._replace(outside=True,
-                        frame=prev_shape.frame + task_data.frame_step)
+                        frame=prev_shape.frame + instance_data.frame_step)
                 prev_shape_idx += 1
                 track.shapes.insert(prev_shape_idx, prev_shape)
             prev_shape = shape
@@ -78,12 +81,12 @@ def _import_task(dataset, task_data):
 
         # Append a shape with outside=True to finish the track
         last_shape = track.shapes[-1]
-        if last_shape.frame + task_data.frame_step <= \
-                int(task_data.meta['task']['stop_frame']):
+        if last_shape.frame + instance_data.frame_step <= \
+                int(instance_data.meta[instance_data.META_FIELD]['stop_frame']):
             track.shapes.append(last_shape._replace(outside=True,
-                frame=last_shape.frame + task_data.frame_step)
+                frame=last_shape.frame + instance_data.frame_step)
             )
-        task_data.add_track(track)
+        instance_data.add_track(track)
 
 
 @exporter(name='MOT', ext='ZIP', version='1.1')
@@ -107,7 +110,7 @@ def _import(src_file, instance_data, load_data_callback=None):
         # Dirty way to determine instance type to avoid circular dependency
         if hasattr(instance_data, '_db_project'):
             for sub_dataset, task_data in instance_data.split_dataset(dataset):
-                _import_task(sub_dataset, task_data)
+                _import_to_task(sub_dataset, task_data)
         else:
-            _import_task(dataset, instance_data)
+            _import_to_task(dataset, instance_data)
 

cvat/apps/dataset_manager/formats/mots.py

 # Copyright (C) 2019-2022 Intel Corporation
+# Copyright (C) 2022 CVAT.ai Corporation
 #
 # SPDX-License-Identifier: MIT
 
@@ -22,16 +23,16 @@ class KeepTracks(ItemTransform):
         return item.wrap(annotations=[a for a in item.annotations
             if 'track_id' in a.attributes])
 
-def _import_task(dataset, task_data):
+def _import_to_task(dataset, instance_data):
     tracks = {}
     label_cat = dataset.categories()[AnnotationType.label]
 
-    root_hint = find_dataset_root(dataset, task_data)
+    root_hint = find_dataset_root(dataset, instance_data)
 
     shift = 0
     for item in dataset:
-        frame_number = task_data.abs_frame_id(
-            match_dm_item(item, task_data, root_hint=root_hint))
+        frame_number = instance_data.abs_frame_id(
+            match_dm_item(item, instance_data, root_hint=root_hint))
 
         track_ids = set()
 
@@ -49,7 +50,7 @@ def _import_task(dataset, task_data):
             else:
                 track_ids.add(track_id)
 
-            shape = task_data.TrackedShape(
+            shape = instance_data.TrackedShape(
                 type='polygon',
                 points=ann.points,
                 occluded=ann.attributes.get('occluded') is True,
@@ -64,7 +65,7 @@ def _import_task(dataset, task_data):
 
             # build trajectories as lists of shapes in track dict
             if track_id not in tracks:
-                tracks[track_id] = task_data.Track(
+                tracks[track_id] = instance_data.Track(
                     label_cat.items[ann.label].name, 0, 'manual', [])
             tracks[track_id].shapes.append(shape)
 
@@ -75,10 +76,10 @@ def _import_task(dataset, task_data):
         prev_shape_idx = 0
         prev_shape = track.shapes[0]
         for shape in track.shapes[1:]:
-            has_skip = task_data.frame_step < shape.frame - prev_shape.frame
+            has_skip = instance_data.frame_step < shape.frame - prev_shape.frame
             if has_skip and not prev_shape.outside:
                 prev_shape = prev_shape._replace(outside=True,
-                        frame=prev_shape.frame + task_data.frame_step)
+                        frame=prev_shape.frame + instance_data.frame_step)
                 prev_shape_idx += 1
                 track.shapes.insert(prev_shape_idx, prev_shape)
             prev_shape = shape
@@ -86,12 +87,12 @@ def _import_task(dataset, task_data):
 
         # Append a shape with outside=True to finish the track
         last_shape = track.shapes[-1]
-        if last_shape.frame + task_data.frame_step <= \
-                int(task_data.meta['task']['stop_frame']):
+        if last_shape.frame + instance_data.frame_step <= \
+                int(instance_data.meta[instance_data.META_FIELD]['stop_frame']):
             track.shapes.append(last_shape._replace(outside=True,
-                frame=last_shape.frame + task_data.frame_step)
+                frame=last_shape.frame + instance_data.frame_step)
             )
-        task_data.add_track(track)
+        instance_data.add_track(track)
 
 @exporter(name='MOTS PNG', ext='ZIP', version='1.0')
 def _export(dst_file, instance_data, save_images=False):
@@ -120,7 +121,7 @@ def _import(src_file, instance_data, load_data_callback=None):
         # Dirty way to determine instance type to avoid circular dependency
         if hasattr(instance_data, '_db_project'):
             for sub_dataset, task_data in instance_data.split_dataset(dataset):
-                _import_task(sub_dataset, task_data)
+                _import_to_task(sub_dataset, task_data)
         else:
-            _import_task(dataset, instance_data)
+            _import_to_task(dataset, instance_data)
 

cvat/apps/dataset_manager/formats/pascal_voc.py

 # Copyright (C) 2020-2022 Intel Corporation
+# Copyright (C) 2022 CVAT.ai Corporation
 #
 # SPDX-License-Identifier: MIT
 
@@ -11,8 +12,7 @@ from tempfile import TemporaryDirectory
 from datumaro.components.dataset import Dataset
 from pyunpack import Archive
 
-from cvat.apps.dataset_manager.bindings import (GetCVATDataExtractor,
-    ProjectData, import_dm_annotations)
+from cvat.apps.dataset_manager.bindings import (GetCVATDataExtractor, import_dm_annotations)
 from cvat.apps.dataset_manager.util import make_zip_archive
 
 from .registry import dm_env, exporter, importer
@@ -36,8 +36,7 @@ def _import(src_file, instance_data, load_data_callback=None):
         # put label map from the task if not present
         labelmap_file = osp.join(tmp_dir, 'labelmap.txt')
         if not osp.isfile(labelmap_file):
-            labels_meta = instance_data.meta['project']['labels'] \
-                if isinstance(instance_data, ProjectData) else instance_data.meta['task']['labels']
+            labels_meta = instance_data.meta[instance_data.META_FIELD]['labels']
             labels = (label['name'] + ':::' for _, label in labels_meta)
             with open(labelmap_file, 'w') as f:
                 f.write('\n'.join(labels))

cvat/apps/dataset_manager/formats/utils.py

@@ -49,8 +49,7 @@ def hex2rgb(color):
     return tuple(int(color.lstrip('#')[i:i+2], 16) for i in (0, 2, 4))
 
 def make_colormap(instance_data):
-    instance_name = 'project' if 'project' in instance_data.meta.keys() else 'task'
-    labels = [label for _, label in instance_data.meta[instance_name]['labels']]
+    labels = [label for _, label in instance_data.meta[instance_data.META_FIELD]['labels']]
     label_names = [label['name'] for label in labels]
 
     if 'background' not in label_names:

cvat/apps/dataset_manager/task.py

@@ -16,7 +16,7 @@ from cvat.apps.engine.plugins import plugin_decorator
 from cvat.apps.profiler import silk_profile
 
 from .annotation import AnnotationIR, AnnotationManager
-from .bindings import TaskData
+from .bindings import TaskData, JobData
 from .formats.registry import make_exporter, make_importer
 from .util import bulk_create
 
@@ -553,24 +553,24 @@ class JobAnnotation:
         return self.ir_data.data
 
     def export(self, dst_file, exporter, host='', **options):
-        task_data = TaskData(
+        job_data = JobData(
             annotation_ir=self.ir_data,
-            db_task=self.db_job.segment.task,
+            db_job=self.db_job,
             host=host,
         )
-        exporter(dst_file, task_data, **options)
+        exporter(dst_file, job_data, **options)
 
     def import_annotations(self, src_file, importer):
-        task_data = TaskData(
+        job_data = JobData(
             annotation_ir=AnnotationIR(),
-            db_task=self.db_job.segment.task,
+            db_job=self.db_job,
             create_callback=self.create,
         )
         self.delete()
 
-        importer(src_file, task_data)
+        importer(src_file, job_data)
 
-        self.create(task_data.data.slice(self.start_frame, self.stop_frame).serialize())
+        self.create(job_data.data.slice(self.start_frame, self.stop_frame).serialize())
 
 class TaskAnnotation:
     def __init__(self, pk):

cvat/apps/dataset_manager/tests/test_formats.py

@@ -21,7 +21,7 @@ from rest_framework.test import APIClient, APITestCase
 
 import cvat.apps.dataset_manager as dm
 from cvat.apps.dataset_manager.annotation import AnnotationIR
-from cvat.apps.dataset_manager.bindings import (CvatTaskDataExtractor,
+from cvat.apps.dataset_manager.bindings import (CvatTaskOrJobDataExtractor,
                                                 TaskData, find_dataset_root)
 from cvat.apps.dataset_manager.task import TaskAnnotation
 from cvat.apps.dataset_manager.util import make_zip_archive
@@ -417,7 +417,7 @@ class TaskExportTest(_DbTestBase):
         task_ann.init_from_db()
         task_data = TaskData(task_ann.ir_data, Task.objects.get(pk=task["id"]))
 
-        extractor = CvatTaskDataExtractor(task_data)
+        extractor = CvatTaskOrJobDataExtractor(task_data)
         dm_dataset = datumaro.components.project.Dataset.from_extractors(extractor)
         self.assertEqual(4, len(dm_dataset.get("image_1").annotations))
 

cvat/apps/dataset_manager/tests/test_rest_api_formats.py

 # Copyright (C) 2021-2022 Intel Corporation
+# Copyright (C) 2022 CVAT.ai Corporation
 #
 # SPDX-License-Identifier: MIT
 
@@ -22,7 +23,7 @@ from rest_framework import status
 from rest_framework.test import APIClient, APITestCase
 
 import cvat.apps.dataset_manager as dm
-from cvat.apps.dataset_manager.bindings import CvatTaskDataExtractor, TaskData
+from cvat.apps.dataset_manager.bindings import CvatTaskOrJobDataExtractor, TaskData
 from cvat.apps.dataset_manager.task import TaskAnnotation
 from cvat.apps.engine.models import Task
 
@@ -185,7 +186,7 @@ class _DbTestBase(APITestCase):
         task_ann = TaskAnnotation(task_id)
         task_ann.init_from_db()
         task_data = TaskData(task_ann.ir_data, Task.objects.get(pk=task_id))
-        extractor = CvatTaskDataExtractor(task_data, include_images=include_images)
+        extractor = CvatTaskOrJobDataExtractor(task_data, include_images=include_images)
         return Dataset.from_extractors(extractor)
 
     def _get_request_with_data(self, path, data, user):

cvat/apps/engine/frame_provider.py

 # Copyright (C) 2020-2022 Intel Corporation
+# Copyright (C) 2022 CVAT.ai Corporation
 #
 # SPDX-License-Identifier: MIT
 
@@ -182,6 +183,6 @@ class FrameProvider:
             return (frame, self.VIDEO_FRAME_MIME)
         return (frame, mimetypes.guess_type(frame_name)[0])
 
-    def get_frames(self, quality=Quality.ORIGINAL, out_type=Type.BUFFER):
-        for idx in range(self._db_data.size):
+    def get_frames(self, start_frame, stop_frame, quality=Quality.ORIGINAL, out_type=Type.BUFFER):
+        for idx in range(start_frame, stop_frame):
             yield self.get_frame(idx, quality=quality, out_type=out_type)