Drop support of the TFRecord format

README.md

@@ -155,7 +155,6 @@ For more information about the supported formats, see:
 | [YOLO](https://pjreddie.com/darknet/yolo/)                                                       | ✔️     | ✔️     |
 | [MS COCO Object Detection](http://cocodataset.org/#format-data)                                  | ✔️     | ✔️     |
 | [MS COCO Keypoints Detection](http://cocodataset.org/#format-data)                               | ✔️     | ✔️     |
-| [TFrecord](https://www.tensorflow.org/tutorials/load_data/tfrecord)                              | ✔️     | ✔️     |
 | [MOT](https://motchallenge.net/)                                                                 | ✔️     | ✔️     |
 | [MOTS PNG](https://www.vision.rwth-aachen.de/page/mots)                                          | ✔️     | ✔️     |
 | [LabelMe 3.0](http://labelme.csail.mit.edu/Release3.0)                                           | ✔️     | ✔️     |

changelog.d/20240131_180300_roman_remove_tfrecord.md

@@ -0,0 +1,4 @@
+### Removed
+
+- Removed support for the TFRecord dataset format
+  (<https://github.com/opencv/cvat/pull/7416>)

changelog.d/20240213_010700_roman_import_export_events.md

@@ -0,0 +1,5 @@
+### Added
+
+- Added `dataset:export` and `dataset:import` events that are logged when
+  the user initiates an export or import of a project, task or job
+  (<https://github.com/opencv/cvat/pull/7476>)

cvat/apps/dataset_manager/formats/registry.py

@@ -95,13 +95,6 @@ def make_exporter(name):
     return EXPORT_FORMATS[name]()
 
 
-# Add checking for TF availability to avoid CVAT sever instance / interpreter
-# crash and provide a meaningful diagnostic message in the case of AVX
-# instructions unavailability:
-# https://github.com/openvinotoolkit/cvat/pull/1567
-import datumaro.util.tf_util as TF
-TF.enable_tf_check = True
-
 # pylint: disable=unused-import
 import cvat.apps.dataset_manager.formats.coco
 import cvat.apps.dataset_manager.formats.cvat
@@ -111,7 +104,6 @@ def make_exporter(name):
 import cvat.apps.dataset_manager.formats.mot
 import cvat.apps.dataset_manager.formats.mots
 import cvat.apps.dataset_manager.formats.pascal_voc
-import cvat.apps.dataset_manager.formats.tfrecord
 import cvat.apps.dataset_manager.formats.yolo
 import cvat.apps.dataset_manager.formats.imagenet
 import cvat.apps.dataset_manager.formats.camvid

cvat/apps/dataset_manager/tests/assets/annotations.json

@@ -701,24 +701,6 @@
     ],
     "tracks": []
   },
-  "TFRecord 1.0": {
-    "version": 0,
-    "tags": [],
-    "shapes": [
-      {
-        "type": "rectangle",
-        "occluded": false,
-        "z_order": 0,
-        "points": [16.5, 17.2, 38.89, 25.63],
-        "frame": 0,
-        "label_id": null,
-        "group": 0,
-        "source": "manual",
-        "attributes": []
-      }
-    ],
-    "tracks": []
-  },
   "YOLO 1.1": {
     "version": 0,
     "tags": [],

cvat/apps/dataset_manager/tests/test_formats.py

@@ -295,7 +295,6 @@ def test_export_formats_query(self):
             'MOTS PNG 1.0',
             'PASCAL VOC 1.1',
             'Segmentation mask 1.1',
-            'TFRecord 1.0',
             'YOLO 1.1',
             'ImageNet 1.0',
             'CamVid 1.0',
@@ -326,7 +325,6 @@ def test_import_formats_query(self):
             'MOTS PNG 1.0',
             'PASCAL VOC 1.1',
             'Segmentation mask 1.1',
-            'TFRecord 1.0',
             'YOLO 1.1',
             'ImageNet 1.0',
             'CamVid 1.0',
@@ -381,7 +379,6 @@ def test_empty_images_are_exported(self):
             # ('MOTS PNG 1.0', 'mots_png'), # does not support
             ('PASCAL VOC 1.1', 'voc'),
             ('Segmentation mask 1.1', 'voc'),
-            ('TFRecord 1.0', 'tf_detection_api'),
             ('YOLO 1.1', 'yolo'),
             ('ImageNet 1.0', 'imagenet_txt'),
             ('CamVid 1.0', 'camvid'),

cvat/apps/dataset_manager/tests/test_rest_api_formats.py

@@ -417,7 +417,7 @@ def test_api_v2_dump_and_upload_annotations_with_objects_type_is_shape(self):
                         "Cityscapes 1.0", "Datumaro 1.0",
                         "ImageNet 1.0", "MOTS PNG 1.0",
                         "PASCAL VOC 1.1", "Segmentation mask 1.1",
-                        "TFRecord 1.0", "VGGFace2 1.0",
+                        "VGGFace2 1.0",
                         "WiderFace 1.0", "YOLO 1.1"
                     ]:
                         self._create_annotations(task, dump_format_name, "default")
@@ -522,7 +522,7 @@ def test_api_v2_dump_annotations_with_objects_type_is_track(self):
                     if dump_format_name in [
                             "Cityscapes 1.0", "ImageNet 1.0",
                             "MOTS PNG 1.0", "PASCAL VOC 1.1",
-                            "Segmentation mask 1.1", "TFRecord 1.0",
+                            "Segmentation mask 1.1",
                             "VGGFace2 1.0", "WiderFace 1.0", "YOLO 1.1"
                     ]:
                         self._create_annotations(task, dump_format_name, "default")
@@ -968,7 +968,7 @@ def test_api_v2_rewriting_annotations(self):
 
                     if dump_format_name in [
                         "MOT 1.1", "PASCAL VOC 1.1", "Segmentation mask 1.1",
-                        "TFRecord 1.0", "YOLO 1.1", "ImageNet 1.0",
+                        "YOLO 1.1", "ImageNet 1.0",
                         "WiderFace 1.0", "VGGFace2 1.0",
                         "Datumaro 1.0", "Open Images V6 1.0", "KITTI 1.0"
                     ]:
@@ -1084,7 +1084,7 @@ def test_api_v2_tasks_annotations_dump_and_upload_with_datumaro(self):
                     if dump_format_name in [
                         "MOT 1.1", "MOTS PNG 1.0",
                         "PASCAL VOC 1.1", "Segmentation mask 1.1",
-                        "TFRecord 1.0", "YOLO 1.1", "ImageNet 1.0",
+                        "YOLO 1.1", "ImageNet 1.0",
                         "WiderFace 1.0", "VGGFace2 1.0", "LFW 1.0",
                         "Open Images V6 1.0", "Datumaro 1.0", "KITTI 1.0"
                     ]:
@@ -1287,7 +1287,7 @@ def test_api_v2_export_import_dataset(self):
                 dump_format_name = dump_format.DISPLAY_NAME
                 if dump_format_name in [
                     'Cityscapes 1.0', 'LFW 1.0', 'Market-1501 1.0',
-                    'MOT 1.1', 'TFRecord 1.0'
+                    'MOT 1.1',
                 ]:
                     # TO-DO: fix bug for this formats
                     continue
@@ -1303,7 +1303,7 @@ def test_api_v2_export_import_dataset(self):
                 if dump_format_name in [
                     "Cityscapes 1.0", "Datumaro 1.0", "ImageNet 1.0",
                     "MOT 1.1", "MOTS PNG 1.0", "PASCAL VOC 1.1",
-                    "Segmentation mask 1.1", "TFRecord 1.0", "VGGFace2 1.0",
+                    "Segmentation mask 1.1", "VGGFace2 1.0",
                     "WiderFace 1.0", "YOLO 1.1"
                 ]:
                     self._create_annotations(task, dump_format_name, "default")
@@ -1344,7 +1344,7 @@ def test_api_v2_export_import_dataset(self):
                 upload_format_name = upload_format.DISPLAY_NAME
                 if upload_format_name in [
                     'Cityscapes 1.0', 'LFW 1.0', 'Market-1501 1.0',
-                    'MOT 1.1', 'TFRecord 1.0'
+                    'MOT 1.1',
                 ]:
                     # TO-DO: fix bug for this formats
                     continue

cvat/apps/engine/tests/test_rest_api.py

@@ -6031,8 +6031,7 @@ def _get_initial_annotation(annotation_format):
                 annotations["shapes"] = rectangle_shapes_wo_attrs
                 annotations["tags"] = tags_wo_attrs
 
-            elif annotation_format == "YOLO 1.1" or \
-                 annotation_format == "TFRecord 1.0":
+            elif annotation_format == "YOLO 1.1":
                 annotations["shapes"] = rectangle_shapes_wo_attrs
 
             elif annotation_format == "COCO 1.0":
@@ -6404,8 +6403,6 @@ def etree_to_dict(t):
                 for json in jsons:
                     coco = coco_loader.COCO(json)
                     self.assertTrue(coco.getAnnIds())
-        elif format_name == "TFRecord 1.0":
-            self.assertTrue(zipfile.is_zipfile(content))
         elif format_name == "Segmentation mask 1.1":
             self.assertTrue(zipfile.is_zipfile(content))
 

cvat/apps/engine/views.py

@@ -43,6 +43,7 @@
 import cvat.apps.dataset_manager as dm
 import cvat.apps.dataset_manager.views  # pylint: disable=unused-import
 from cvat.apps.engine.cloud_provider import db_storage_to_storage_instance, download_file_from_bucket, export_resource_to_cloud_storage
+from cvat.apps.events.handlers import handle_dataset_export, handle_dataset_import
 from cvat.apps.dataset_manager.bindings import CvatImportError
 from cvat.apps.dataset_manager.serializers import DatasetFormatsSerializer
 from cvat.apps.engine.frame_provider import FrameProvider
@@ -2838,6 +2839,8 @@ def _import_annotations(request, rq_id_template, rq_func, db_obj, format_name,
         dependent_job = None
         location = location_conf.get('location') if location_conf else Location.LOCAL
 
+        db_storage = None
+
         if not filename or location == Location.CLOUD_STORAGE:
             if location != Location.CLOUD_STORAGE:
                 serializer = AnnotationFileSerializer(data=request.data)
@@ -2898,6 +2901,9 @@ def _import_annotations(request, rq_id_template, rq_func, db_obj, format_name,
                 result_ttl=settings.IMPORT_CACHE_SUCCESS_TTL.total_seconds(),
                 failure_ttl=settings.IMPORT_CACHE_FAILED_TTL.total_seconds()
             )
+
+        handle_dataset_import(db_obj, format_name=format_name, cloud_storage=db_storage)
+
         serializer = RqIdSerializer(data={'rq_id': rq_id})
         serializer.is_valid(raise_exception=True)
 
@@ -3045,6 +3051,8 @@ def _export_annotations(
             is_annotation_file=is_annotation_file,
         )
         func_args = (db_storage, filename, filename_pattern, callback) + func_args
+    else:
+        db_storage = None
 
     with get_rq_lock_by_user(queue, user_id):
         queue.enqueue_call(
@@ -3056,6 +3064,10 @@ def _export_annotations(
             result_ttl=ttl,
             failure_ttl=ttl,
         )
+
+    handle_dataset_export(db_instance,
+        format_name=format_name, cloud_storage=db_storage, save_images=not is_annotation_file)
+
     return Response(status=status.HTTP_202_ACCEPTED)
 
 def _import_project_dataset(request, rq_id_template, rq_func, db_obj, format_name, filename=None, conv_mask_to_poly=True, location_conf=None):
@@ -3080,6 +3092,8 @@ def _import_project_dataset(request, rq_id_template, rq_func, db_obj, format_nam
             rq_job.delete()
         dependent_job = None
         location = location_conf.get('location') if location_conf else None
+        db_storage = None
+
         if not filename and location != Location.CLOUD_STORAGE:
             serializer = DatasetFileSerializer(data=request.data)
             if serializer.is_valid(raise_exception=True):
@@ -3138,6 +3152,8 @@ def _import_project_dataset(request, rq_id_template, rq_func, db_obj, format_nam
                 result_ttl=settings.IMPORT_CACHE_SUCCESS_TTL.total_seconds(),
                 failure_ttl=settings.IMPORT_CACHE_FAILED_TTL.total_seconds()
             )
+
+        handle_dataset_import(db_obj, format_name=format_name, cloud_storage=db_storage)
     else:
         return Response(status=status.HTTP_409_CONFLICT, data='Import job already exists')
 

cvat/apps/events/event.py

@@ -26,6 +26,7 @@ class EventScopes:
         "comment": ["create", "update", "delete"],
         "annotations": ["create", "update", "delete"],
         "label": ["create", "update", "delete"],
+        "dataset": ["export", "import"],
     }
 
     @classmethod

cvat/apps/events/handlers.py

@@ -3,6 +3,7 @@
 # SPDX-License-Identifier: MIT
 
 from copy import deepcopy
+from typing import Optional, Union
 import traceback
 import rq
 
@@ -474,6 +475,51 @@ def filter_shape_data(shape):
                 payload={"tracks": tracks},
             )
 
+def handle_dataset_io(
+    instance: Union[Project, Task, Job],
+    action: str,
+    *,
+    format_name: str,
+    cloud_storage: Optional[CloudStorage],
+    **payload_fields,
+) -> None:
+    payload={"format": format_name, **payload_fields}
+
+    if cloud_storage:
+        payload["cloud_storage"] = {"id": cloud_storage.id}
+
+    record_server_event(
+        scope=event_scope(action, "dataset"),
+        request_id=request_id(),
+        org_id=organization_id(instance),
+        org_slug=organization_slug(instance),
+        project_id=project_id(instance),
+        task_id=task_id(instance),
+        job_id=job_id(instance),
+        user_id=user_id(instance),
+        user_name=user_name(instance),
+        user_email=user_email(instance),
+        payload=payload,
+    )
+
+def handle_dataset_export(
+    instance: Union[Project, Task, Job],
+    *,
+    format_name: str,
+    cloud_storage: Optional[CloudStorage],
+    save_images: bool,
+) -> None:
+    handle_dataset_io(instance, "export",
+        format_name=format_name, cloud_storage=cloud_storage, save_images=save_images)
+
+def handle_dataset_import(
+    instance: Union[Project, Task, Job],
+    *,
+    format_name: str,
+    cloud_storage: Optional[CloudStorage],
+) -> None:
+    handle_dataset_io(instance, "import", format_name=format_name, cloud_storage=cloud_storage)
+
 def handle_rq_exception(rq_job, exc_type, exc_value, tb):
     oid = rq_job.meta.get("org_id", None)
     oslug = rq_job.meta.get("org_slug", None)

cvat/requirements/base.in

@@ -51,4 +51,3 @@ rq-scheduler==0.13.1
 rq==1.15.1
 rules>=3.3
 Shapely==1.7.1
-tensorflow==2.11.1 # Optional requirement of Datumaro. Use tensorflow-macos==2.8.0 for Mac M1