|
| 1 | +from datetime import datetime |
| 2 | +from unittest.mock import Mock, PropertyMock |
| 3 | + |
| 4 | +import numpy as np |
| 5 | +import pytest |
| 6 | + |
| 7 | +from encord.objects import LabelRowV2, Object, ObjectInstance, Shape |
| 8 | +from encord.objects.bitmask import BitmaskCoordinates |
| 9 | +from encord.orm.label_row import AnnotationTaskStatus, LabelRowMetadata, LabelStatus |
| 10 | + |
| 11 | +bitmask_object = Object( |
| 12 | + uid=1, name="Mask", color="#D33115", shape=Shape.BITMASK, feature_node_hash="bitmask123", attributes=[] |
| 13 | +) |
| 14 | +get_child_by_hash = PropertyMock(return_value=bitmask_object) |
| 15 | +ontology_structure = Mock(get_child_by_hash=get_child_by_hash) |
| 16 | +ontology = Mock(structure=ontology_structure) |
| 17 | + |
| 18 | + |
| 19 | +def test_image_bitmask_dimension_validation(): |
| 20 | + # Create label row metadata with specific dimensions (512x512) |
| 21 | + metadata = LabelRowMetadata( |
| 22 | + label_hash="test_label", |
| 23 | + branch_name="main", |
| 24 | + created_at=datetime.now(), |
| 25 | + last_edited_at=datetime.now(), |
| 26 | + data_hash="test_data", |
| 27 | + data_title="Test Image", |
| 28 | + data_type="IMAGE", |
| 29 | + data_link="", |
| 30 | + dataset_hash="test_dataset", |
| 31 | + dataset_title="Test Dataset", |
| 32 | + label_status=LabelStatus.NOT_LABELLED, |
| 33 | + annotation_task_status=AnnotationTaskStatus.QUEUED, |
| 34 | + workflow_graph_node=None, |
| 35 | + is_shadow_data=False, |
| 36 | + duration=None, |
| 37 | + frames_per_second=None, |
| 38 | + number_of_frames=1, |
| 39 | + height=512, |
| 40 | + width=512, |
| 41 | + audio_codec=None, |
| 42 | + audio_sample_rate=None, |
| 43 | + audio_num_channels=None, |
| 44 | + audio_bit_depth=None, |
| 45 | + ) |
| 46 | + |
| 47 | + # Create empty labels dict |
| 48 | + empty_labels = { |
| 49 | + "label_hash": "test_label", |
| 50 | + "branch_name": "main", |
| 51 | + "created_at": "Thu, 09 Feb 2023 14:12:03 UTC", |
| 52 | + "last_edited_at": "Thu, 09 Feb 2023 14:12:03 UTC", |
| 53 | + "data_hash": "test_data", |
| 54 | + "annotation_task_status": "QUEUED", |
| 55 | + "is_shadow_data": False, |
| 56 | + "dataset_hash": "test_dataset", |
| 57 | + "dataset_title": "Test Dataset", |
| 58 | + "data_title": "Test Image", |
| 59 | + "data_type": "image", |
| 60 | + "data_units": { |
| 61 | + "test_data": { |
| 62 | + "data_hash": "test_data", |
| 63 | + "data_title": "Test Image", |
| 64 | + "data_link": "", |
| 65 | + "data_type": "image/png", |
| 66 | + "data_sequence": "0", |
| 67 | + "width": 512, |
| 68 | + "height": 512, |
| 69 | + "labels": {"objects": [], "classifications": []}, |
| 70 | + } |
| 71 | + }, |
| 72 | + "object_answers": {}, |
| 73 | + "classification_answers": {}, |
| 74 | + "object_actions": {}, |
| 75 | + "label_status": "LABEL_IN_PROGRESS", |
| 76 | + } |
| 77 | + |
| 78 | + # Correct dimensions (512x512) should succeed |
| 79 | + label_row = LabelRowV2(metadata, Mock(), ontology) |
| 80 | + label_row.from_labels_dict(empty_labels) |
| 81 | + |
| 82 | + correct_bitmask_coords = BitmaskCoordinates(np.zeros((512, 512), dtype=bool)) |
| 83 | + correct_obj_instance = ObjectInstance(bitmask_object) |
| 84 | + correct_obj_instance.set_for_frames(coordinates=correct_bitmask_coords, frames=0) |
| 85 | + label_row.add_object_instance(correct_obj_instance) |
| 86 | + assert len(label_row.get_object_instances()) == 1 |
| 87 | + label_row.to_encord_dict() # Serialization should also succeed |
| 88 | + |
| 89 | + # Incorrect dimensions (256x256) should raise ValueError |
| 90 | + incorrect_bitmask_coords = BitmaskCoordinates(np.zeros((256, 256), dtype=bool)) |
| 91 | + incorrect_obj_instance = ObjectInstance(bitmask_object) |
| 92 | + incorrect_obj_instance.set_for_frames(coordinates=incorrect_bitmask_coords, frames=0) |
| 93 | + label_row.add_object_instance(incorrect_obj_instance) |
| 94 | + assert len(label_row.get_object_instances()) == 2 |
| 95 | + |
| 96 | + with pytest.raises(ValueError, match="Bitmask dimensions don't match the media dimensions"): |
| 97 | + label_row.to_encord_dict() |
| 98 | + |
| 99 | + |
| 100 | +def test_image_group_bitmask_dimension_validation(): |
| 101 | + metadata = LabelRowMetadata( |
| 102 | + label_hash="test_label", |
| 103 | + branch_name="main", |
| 104 | + created_at=datetime.now(), |
| 105 | + last_edited_at=datetime.now(), |
| 106 | + data_hash="test_data", |
| 107 | + data_title="Test Image Group", |
| 108 | + data_type="IMG_GROUP", |
| 109 | + data_link="", |
| 110 | + dataset_hash="test_dataset", |
| 111 | + dataset_title="Test Dataset", |
| 112 | + label_status=LabelStatus.NOT_LABELLED, |
| 113 | + annotation_task_status=AnnotationTaskStatus.QUEUED, |
| 114 | + workflow_graph_node=None, |
| 115 | + is_shadow_data=False, |
| 116 | + duration=None, |
| 117 | + frames_per_second=None, |
| 118 | + number_of_frames=2, |
| 119 | + height=None, |
| 120 | + width=None, |
| 121 | + audio_codec=None, |
| 122 | + audio_sample_rate=None, |
| 123 | + audio_num_channels=None, |
| 124 | + audio_bit_depth=None, |
| 125 | + ) |
| 126 | + |
| 127 | + # Create empty labels dict for image group with different dimensions per frame |
| 128 | + empty_labels = { |
| 129 | + "label_hash": "test_label", |
| 130 | + "branch_name": "main", |
| 131 | + "created_at": "Thu, 09 Feb 2023 14:12:03 UTC", |
| 132 | + "last_edited_at": "Thu, 09 Feb 2023 14:12:03 UTC", |
| 133 | + "data_hash": "test_data", |
| 134 | + "annotation_task_status": "QUEUED", |
| 135 | + "is_shadow_data": False, |
| 136 | + "dataset_hash": "test_dataset", |
| 137 | + "dataset_title": "Test Dataset", |
| 138 | + "data_title": "Test Image Group", |
| 139 | + "data_type": "img_group", |
| 140 | + "data_units": { |
| 141 | + "frame_0_hash": { |
| 142 | + "data_hash": "frame_0_hash", |
| 143 | + "data_title": "Frame 0", |
| 144 | + "data_link": "", |
| 145 | + "data_type": "image/png", |
| 146 | + "data_sequence": "0", |
| 147 | + "width": 512, |
| 148 | + "height": 512, |
| 149 | + "labels": {"objects": [], "classifications": []}, |
| 150 | + }, |
| 151 | + "frame_1_hash": { |
| 152 | + "data_hash": "frame_1_hash", |
| 153 | + "data_title": "Frame 1", |
| 154 | + "data_link": "", |
| 155 | + "data_type": "image/png", |
| 156 | + "data_sequence": "1", |
| 157 | + "width": 1024, |
| 158 | + "height": 768, |
| 159 | + "labels": {"objects": [], "classifications": []}, |
| 160 | + }, |
| 161 | + }, |
| 162 | + "object_answers": {}, |
| 163 | + "classification_answers": {}, |
| 164 | + "object_actions": {}, |
| 165 | + "label_status": "LABEL_IN_PROGRESS", |
| 166 | + } |
| 167 | + |
| 168 | + label_row = LabelRowV2(metadata, Mock(), ontology) |
| 169 | + label_row.from_labels_dict(empty_labels) |
| 170 | + |
| 171 | + # Add bitmask with correct dimensions for frame 0 (512x512) |
| 172 | + frame_0_correct_mask = BitmaskCoordinates(np.zeros((512, 512), dtype=bool)) |
| 173 | + frame_0_correct_instance = ObjectInstance(bitmask_object) |
| 174 | + frame_0_correct_instance.set_for_frames(coordinates=frame_0_correct_mask, frames=0) |
| 175 | + label_row.add_object_instance(frame_0_correct_instance) |
| 176 | + |
| 177 | + # Add bitmask with correct dimensions for frame 1 (1024x768) |
| 178 | + frame_1_correct_mask = BitmaskCoordinates(np.zeros((768, 1024), dtype=bool)) |
| 179 | + frame_1_correct_instance = ObjectInstance(bitmask_object) |
| 180 | + frame_1_correct_instance.set_for_frames(coordinates=frame_1_correct_mask, frames=1) |
| 181 | + label_row.add_object_instance(frame_1_correct_instance) |
| 182 | + assert len(label_row.get_object_instances()) == 2 |
| 183 | + label_row.to_encord_dict() # Both correct dimensions should serialize successfully |
| 184 | + |
| 185 | + # Add bitmask with incorrect dimensions for frame 0 (256x256 instead of 512x512) |
| 186 | + frame_0_incorrect_mask = BitmaskCoordinates(np.zeros((256, 256), dtype=bool)) |
| 187 | + frame_0_incorrect_instance = ObjectInstance(bitmask_object) |
| 188 | + frame_0_incorrect_instance.set_for_frames(coordinates=frame_0_incorrect_mask, frames=0, overwrite=True) |
| 189 | + label_row.add_object_instance(frame_0_incorrect_instance, force=True) |
| 190 | + |
| 191 | + # Should fail on serialization |
| 192 | + with pytest.raises(ValueError, match="Bitmask dimensions don't match the media dimensions"): |
| 193 | + label_row.to_encord_dict() |
| 194 | + |
| 195 | + # Fix frame 0, then add incorrect dimensions for frame 1 |
| 196 | + label_row.remove_object(frame_0_incorrect_instance) |
| 197 | + label_row.to_encord_dict() # back to a valid serializable state |
| 198 | + |
| 199 | + frame_1_incorrect_mask = BitmaskCoordinates(np.zeros((512, 512), dtype=bool)) |
| 200 | + frame_1_incorrect_instance = ObjectInstance(bitmask_object) |
| 201 | + frame_1_incorrect_instance.set_for_frames(coordinates=frame_1_incorrect_mask, frames=1, overwrite=True) |
| 202 | + label_row.add_object_instance(frame_1_incorrect_instance, force=True) |
| 203 | + |
| 204 | + # Should fail on serialization due to frame 1 |
| 205 | + with pytest.raises(ValueError, match="Bitmask dimensions don't match the media dimensions"): |
| 206 | + label_row.to_encord_dict() |
| 207 | + |
| 208 | + |
| 209 | +def test_dicom_bitmask_dimension_validation(): |
| 210 | + # Create label row metadata for DICOM with series-level dimensions (512x512) |
| 211 | + metadata = LabelRowMetadata( |
| 212 | + label_hash="test_label", |
| 213 | + branch_name="main", |
| 214 | + created_at=datetime.now(), |
| 215 | + last_edited_at=datetime.now(), |
| 216 | + data_hash="test_dicom_hash", |
| 217 | + data_title="Test DICOM", |
| 218 | + data_type="DICOM", |
| 219 | + data_link="", |
| 220 | + dataset_hash="test_dataset", |
| 221 | + dataset_title="Test Dataset", |
| 222 | + label_status=LabelStatus.NOT_LABELLED, |
| 223 | + annotation_task_status=AnnotationTaskStatus.QUEUED, |
| 224 | + workflow_graph_node=None, |
| 225 | + is_shadow_data=False, |
| 226 | + duration=None, |
| 227 | + frames_per_second=None, |
| 228 | + number_of_frames=2, |
| 229 | + height=512, |
| 230 | + width=512, |
| 231 | + audio_codec=None, |
| 232 | + audio_sample_rate=None, |
| 233 | + audio_num_channels=None, |
| 234 | + audio_bit_depth=None, |
| 235 | + ) |
| 236 | + |
| 237 | + dicom_labels = { |
| 238 | + "label_hash": "test_label", |
| 239 | + "branch_name": "main", |
| 240 | + "created_at": "Thu, 09 Feb 2023 14:12:03 UTC", |
| 241 | + "last_edited_at": "Thu, 09 Feb 2023 14:12:03 UTC", |
| 242 | + "data_hash": "test_dicom_hash", |
| 243 | + "annotation_task_status": "QUEUED", |
| 244 | + "is_shadow_data": False, |
| 245 | + "dataset_hash": "test_dataset", |
| 246 | + "dataset_title": "Test Dataset", |
| 247 | + "data_title": "Test DICOM", |
| 248 | + "data_type": "dicom", |
| 249 | + "data_units": { |
| 250 | + "test_dicom_hash": { |
| 251 | + "data_hash": "test_dicom_hash", |
| 252 | + "data_title": "Test DICOM", |
| 253 | + "data_type": "application/dicom", |
| 254 | + "data_sequence": 0, |
| 255 | + "labels": { |
| 256 | + "0": { |
| 257 | + "objects": [], |
| 258 | + "classifications": [], |
| 259 | + "metadata": { |
| 260 | + "dicom_instance_uid": "1.2.3.4.5.6.7.8.9.0", |
| 261 | + "multiframe_frame_number": None, |
| 262 | + "file_uri": "test/slice_0", |
| 263 | + "width": 512, |
| 264 | + "height": 512, |
| 265 | + }, |
| 266 | + }, |
| 267 | + "1": { |
| 268 | + "objects": [], |
| 269 | + "classifications": [], |
| 270 | + "metadata": { |
| 271 | + "dicom_instance_uid": "1.2.3.4.5.6.7.8.9.1", |
| 272 | + "multiframe_frame_number": None, |
| 273 | + "file_uri": "test/slice_1", |
| 274 | + "width": 10, |
| 275 | + "height": 10, |
| 276 | + }, |
| 277 | + }, |
| 278 | + }, |
| 279 | + "metadata": { |
| 280 | + "patient_id": "test_patient", |
| 281 | + "study_uid": "1.2.3.4.5", |
| 282 | + "series_uid": "1.2.3.4.6", |
| 283 | + }, |
| 284 | + "data_links": ["test/slice_0", "test/slice_1"], |
| 285 | + "width": 512, |
| 286 | + "height": 512, |
| 287 | + } |
| 288 | + }, |
| 289 | + "object_answers": {}, |
| 290 | + "classification_answers": {}, |
| 291 | + "object_actions": {}, |
| 292 | + "label_status": "LABEL_IN_PROGRESS", |
| 293 | + } |
| 294 | + |
| 295 | + label_row = LabelRowV2(metadata, Mock(), ontology) |
| 296 | + label_row.from_labels_dict(dicom_labels) |
| 297 | + |
| 298 | + slice_0_metadata = label_row.get_frame_view(0).metadata |
| 299 | + assert slice_0_metadata is not None |
| 300 | + assert slice_0_metadata.model_dump() == { |
| 301 | + "width": 512, |
| 302 | + "height": 512, |
| 303 | + "dicom_instance_uid": "1.2.3.4.5.6.7.8.9.0", |
| 304 | + "multiframe_frame_number": None, |
| 305 | + "file_uri": "test/slice_0", |
| 306 | + } |
| 307 | + |
| 308 | + slice_1_metadata = label_row.get_frame_view(1).metadata |
| 309 | + assert slice_1_metadata is not None |
| 310 | + assert slice_1_metadata.model_dump() == { |
| 311 | + "width": 10, |
| 312 | + "height": 10, |
| 313 | + "dicom_instance_uid": "1.2.3.4.5.6.7.8.9.1", |
| 314 | + "multiframe_frame_number": None, |
| 315 | + "file_uri": "test/slice_1", |
| 316 | + } |
| 317 | + |
| 318 | + # Add bitmask with correct dimensions for slice 0 |
| 319 | + slice_0_correct_mask = BitmaskCoordinates(np.zeros((512, 512), dtype=bool)) |
| 320 | + slice_0_correct_instance = ObjectInstance(bitmask_object) |
| 321 | + slice_0_correct_instance.set_for_frames(coordinates=slice_0_correct_mask, frames=0) |
| 322 | + label_row.add_object_instance(slice_0_correct_instance) |
| 323 | + assert len(label_row.get_object_instances()) == 1 |
| 324 | + |
| 325 | + # Add bitmask with correct dimensions for slice 1 |
| 326 | + slice_1_correct_mask = BitmaskCoordinates(np.zeros((10, 10), dtype=bool)) |
| 327 | + slice_1_correct_instance = ObjectInstance(bitmask_object) |
| 328 | + slice_1_correct_instance.set_for_frames(coordinates=slice_1_correct_mask, frames=1) |
| 329 | + label_row.add_object_instance(slice_1_correct_instance) |
| 330 | + assert len(label_row.get_object_instances()) == 2 |
| 331 | + |
| 332 | + # Both correct dimensions should serialize successfully |
| 333 | + label_row.to_encord_dict() |
| 334 | + |
| 335 | + # Add bitmask with incorrect dimensions for slice 0 (256x256 instead of 512x512) |
| 336 | + slice_0_incorrect_mask = BitmaskCoordinates(np.zeros((256, 256), dtype=bool)) |
| 337 | + slice_0_incorrect_instance = ObjectInstance(bitmask_object) |
| 338 | + slice_0_incorrect_instance.set_for_frames(coordinates=slice_0_incorrect_mask, frames=0, overwrite=True) |
| 339 | + label_row.add_object_instance(slice_0_incorrect_instance, force=True) |
| 340 | + |
| 341 | + # Should fail on serialization |
| 342 | + with pytest.raises(ValueError, match="Bitmask dimensions don't match the media dimensions"): |
| 343 | + label_row.to_encord_dict() |
0 commit comments