34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729 | class ApocContainer(Container):
"""
Container class for managing the ApocContainer widget in napari.
Parameters
----------
viewer : napari.viewer.Viewer
The napari viewer instance.
Attributes
----------
_viewer : napari.viewer.Viewer
The napari viewer instance.
_image_directory : FileEdit
Widget for selecting the image directory.
_label_directory : FileEdit
Widget for selecting the label directory.
_output_directory : FileEdit
Widget for selecting the output directory.
_classifier_file : FileEdit
Widget for selecting the classifier file.
_classifier_type_mapping : dict
Mapping of classifier types to their corresponding classes.
_classifier_type : RadioButtons
Widget for selecting the classifier type.
_max_depth : SpinBox
Widget for selecting the number of forests.
_num_trees : SpinBox
Widget for selecting the number of trees.
_positive_class_id : SpinBox
Widget for selecting the object label ID.
_image_channels : Select
Widget for selecting the image channels.
_channel_order_label : Label
Label widget for displaying the selected channel order.
_PDFS : Enum
Enum for predefined feature sets.
_predefined_features : ComboBox
Widget for selecting the features.
_custom_features : LineEdit
Widget for entering custom feature string.
_open_custom_feature_generator : PushButton
Button for opening the custom feature generator widget.
_continue_training : CheckBox
Checkbox for indicating whether to continue training.
_batch_train_button : PushButton
Button for training the classifier on image-label pairs.
_batch_predict_button : PushButton
Button for predicting labels with the classifier.
_progress_bar : ProgressBar
Progress bar widget.
_image_layer : Select
Widget for selecting the image layers.
_label_layer : Widget
Widget for selecting the label layers.
_train_image_button : PushButton
Button for training the classifier on selected layers using labels.
_predict_image_layer : PushButton
Button for predicting using the classifier on selected layers.
_single_result_label : Label
Label widget for displaying a single result.
Methods
-------
_update_metadata_from_file()
Update the metadata from the selected image directory.
_update_channel_order()
Update the channel order label based on the selected image channels.
_set_value_from_pattern(pattern, content)
Set the value from a pattern in the content.
_process_classifier_metadata(content)
Process the classifier metadata from the content.
_update_classifier_metadata()
Update the classifier metadata based on the selected classifier file.
_classifier_statistics_table(custom_classifier)
Display the classifier statistics table.
_get_feature_set()
Get the selected feature set.
_get_training_classifier_instance()
Get the training classifier instance based on the selected classifier
type.
_get_channel_image(img, channel_index_list)
Get the channel image based on the selected channel index list.
"""
def __init__(
self,
viewer: napari.viewer.Viewer = None,
# viewer = napari_viewer
):
super().__init__()
self._viewer = viewer if viewer is not None else None
self._lazy_imports()
self._initialize_widgets()
self._initialize_batch_container()
self._initialize_viewer_container()
self._initialize_custom_apoc_container()
self._setup_widget_layout()
self._connect_events()
def _lazy_imports(self):
import apoc
self.apoc = apoc
def _filter_layers(self, layer_type):
# only do this if the viewer is not None
if self._viewer is None:
return []
return [x for x in self._viewer.layers if isinstance(x, layer_type)]
def _initialize_widgets(self):
self._classifier_file = FileEdit(
label='Classifier File (.cl)',
mode='r',
tooltip='Create a .txt file and rename it to .cl ending.',
)
self._continue_training = CheckBox(
label='Continue Training?',
value=True,
tooltip=(
'Continue training only matters if classifier already exists.'
),
)
self._classifier_type_mapping = {
'PixelClassifier': self.apoc.PixelClassifier,
'ObjectSegmenter': self.apoc.ObjectSegmenter,
}
self._classifier_type = RadioButtons(
label='Classifier Type',
value='ObjectSegmenter',
choices=['ObjectSegmenter', 'PixelClassifier'],
tooltip='Object Segmenter is used for detecting objects of one '
'class, including connected components. '
'Pixel Classifier is used to classify pixel-types.',
)
self._max_depth = SpinBox(
label='Num. of Forests',
value=2,
max=20,
step=1,
tooltip='Increases training time for each forest',
)
self._num_trees = SpinBox(
label='Num. of Trees',
value=100,
step=50,
tooltip='Increases computational requirements.',
)
self._positive_class_id = SpinBox(
label='Object Label ID',
value=2,
step=1,
tooltip='Only used with ObjectSegmenter, otherwise ignored.',
)
self._PDFS = Enum(
'PDFS', self.apoc.PredefinedFeatureSet._member_names_
)
self._predefined_features = ComboBox(
label='Features',
choices=self._PDFS,
nullable=True,
value=None,
tooltip="All featuresets except 'custom' are premade",
)
self._feature_string = LineEdit(
label='Feature String',
tooltip=(
'A string in the form of ' "'filter1=radius1 filter2=radius2'."
),
)
def _initialize_batch_container(self):
self._image_directory = FileEdit(label='Image Directory', mode='d')
self._label_directory = FileEdit(label='Label Directory', mode='d')
self._output_directory = FileEdit(label='Output Directory', mode='d')
self._image_channels = Select(
label='Image Channels',
choices=[],
tooltip=(
'Channel order should be same for training and prediction.'
),
)
self._channel_order_label = Label(value='Select an Image Channel!')
self._batch_train_button = PushButton(label='Train')
self._batch_predict_button = PushButton(label='Predict')
self._batch_train_container = Container(
layout='horizontal',
# label="Train Classifier on Image-Label Pairs",
)
self._batch_train_container.extend(
[self._label_directory, self._batch_train_button]
)
self._batch_predict_container = Container(
layout='horizontal',
# label="Predict Labels with Classifier on Images"
)
self._batch_predict_container.extend(
[self._output_directory, self._batch_predict_button]
)
self._progress_bar = ProgressBar(label='Progress:')
self._batch_container = Container(layout='vertical')
self._batch_container.extend(
[
self._image_directory,
self._image_channels,
self._channel_order_label,
self._batch_train_container,
self._batch_predict_container,
self._progress_bar,
]
)
def _initialize_viewer_container(self):
self._image_layers = Select(
choices=self._filter_layers(layers.Image),
label='Image Layers',
)
self._label_layer = ComboBox(
choices=self._filter_layers(layers.Labels),
label='Label Layer',
)
self._train_image_button = PushButton(
label='Train classifier on selected layers using label'
)
self._predict_image_layer = PushButton(
label='Predict using classifier on selected layers'
)
self._single_result_label = Label()
self._viewer_container = Container(layout='vertical')
self._viewer_container.extend(
[
self._image_layers,
self._label_layer,
self._train_image_button,
self._predict_image_layer,
self._single_result_label,
]
)
def _initialize_custom_apoc_container(self):
from napari_ndev import ApocFeatureStack
self._custom_apoc_container = ApocFeatureStack(viewer=self._viewer)
def _setup_widget_layout(self):
# from napari_ndev import ApocFeatureStack
self.extend(
[
self._classifier_file,
self._continue_training,
self._classifier_type,
self._positive_class_id,
self._max_depth,
self._num_trees,
self._predefined_features,
self._feature_string,
]
)
tabs = QTabWidget()
tabs.addTab(self._batch_container.native, 'Batch')
tabs.addTab(self._viewer_container.native, 'Viewer')
tabs.addTab(self._custom_apoc_container.native, 'Custom Feature Set')
self.native.layout().addWidget(tabs)
def _connect_events(self):
self._image_directory.changed.connect(self._update_metadata_from_file)
self._image_channels.changed.connect(self._update_channel_order)
self._classifier_file.changed.connect(self._update_classifier_metadata)
self._batch_train_button.clicked.connect(self.batch_train)
self._batch_predict_button.clicked.connect(self.batch_predict)
self._train_image_button.clicked.connect(self.image_train)
self._predict_image_layer.clicked.connect(self.image_predict)
self._custom_apoc_container._generate_string_button.clicked.connect(
self.insert_custom_feature_string
)
self._predefined_features.changed.connect(self._get_feature_set)
# when self._viewer.layers is updated, update the choices in the ComboBox
if self._viewer is not None:
self._viewer.layers.events.removed.connect(
self._update_layer_choices
)
self._viewer.layers.events.inserted.connect(
self._update_layer_choices
)
def _update_layer_choices(self):
self._label_layer.choices = self._filter_layers(layers.Labels)
self._image_layers.choices = self._filter_layers(layers.Image)
def _update_metadata_from_file(self):
from aicsimageio import AICSImage
_, files = helpers.get_directory_and_files(self._image_directory.value)
img = AICSImage(files[0])
self._image_channels.choices = helpers.get_channel_names(img)
def _update_channel_order(self):
self._channel_order_label.value = 'Selected Channel Order: ' + str(
self._image_channels.value
)
##############################
# Classifier Related Functions
##############################
def _set_value_from_pattern(self, pattern, content):
match = re.search(pattern, content)
return match.group(1) if match else None
def _process_classifier_metadata(self, content):
self._classifier_type.value = self._set_value_from_pattern(
r'classifier_class_name\s*=\s*([^\n]+)', content
)
self._max_depth.value = self._set_value_from_pattern(
r'max_depth\s*=\s*(\d+)', content
)
self._num_trees.value = self._set_value_from_pattern(
r'num_trees\s*=\s*(\d+)', content
)
self._positive_class_id.value = (
self._set_value_from_pattern(
r'positive_class_identifier\s*=\s*(\d+)', content
)
or 2
)
def _update_classifier_metadata(self):
with open(self._classifier_file.value) as file:
content = file.read()
# Ignore rest of function if file contents are empty
if not content.strip():
return
self._process_classifier_metadata(content)
if self._classifier_type.value in self._classifier_type_mapping:
classifier_class = self._classifier_type_mapping[
self._classifier_type.value
]
custom_classifier = classifier_class(
opencl_filename=self._classifier_file.value
)
else:
custom_classifier = None
self._classifier_statistics_table(custom_classifier)
def _classifier_statistics_table(self, custom_classifier):
table, _ = custom_classifier.statistics()
trans_table = {'filter_name': [], 'radius': []}
for value in table:
filter_name, radius = (
value.split('=') if '=' in value else (value, 0)
)
trans_table['filter_name'].append(filter_name)
trans_table['radius'].append(int(radius))
for i in range(len(next(iter(table.values())))):
trans_table[str(i)] = [round(table[key][i], 2) for key in table]
table_df = pd.DataFrame.from_dict(trans_table)
if self._viewer is not None:
self._viewer.window.add_dock_widget(
Table(value=table_df),
name=os.path.basename(self._classifier_file.value),
)
def _get_feature_set(self):
if self._predefined_features.value.value == 1:
feature_set = ''
else:
feature_set = self.apoc.PredefinedFeatureSet[
self._predefined_features.value.name
].value
self._feature_string.value = feature_set
self._custom_apoc_container._feature_string.value = (
feature_set # <- potentially deprecated in future
)
return feature_set
def _get_training_classifier_instance(self):
if self._classifier_type.value == 'PixelClassifier':
return self.apoc.PixelClassifier(
opencl_filename=self._classifier_file.value,
max_depth=self._max_depth.value,
num_ensembles=self._num_trees.value,
)
if self._classifier_type.value == 'ObjectSegmenter':
return self.apoc.ObjectSegmenter(
opencl_filename=self._classifier_file.value,
positive_class_identifier=self._positive_class_id.value,
max_depth=self._max_depth.value,
num_ensembles=self._num_trees.value,
)
return None
##############################
# Training and Prediction
##############################
def _get_channel_image(self, img, channel_index_list):
if 'S' in img.dims.order:
channel_img = img.get_image_data('TSZYX', S=channel_index_list)
else:
channel_img = img.get_image_data('TCZYX', C=channel_index_list)
return channel_img
def batch_train(self):
from aicsimageio import AICSImage
image_directory, image_files = helpers.get_directory_and_files(
self._image_directory.value
)
label_directory, _ = helpers.get_directory_and_files(
self._label_directory.value
)
# missing_files = check_for_missing_files(image_files, label_directory)
log_loc = self._classifier_file.value.with_suffix('.log.txt')
logger, handler = helpers.setup_logger(log_loc)
logger.info(
"""
Classifier: %s
Channels: %s
Num. Files: %d
Image Directory: %s
Label Directory: %s
""",
self._classifier_file.value,
self._image_channels.value,
len(image_files),
image_directory,
label_directory,
)
# https://github.com/clEsperanto/pyclesperanto_prototype/issues/163
set_wait_for_kernel_finish(True)
self._progress_bar.label = f'Training on %s {len(image_files)} Images'
self._progress_bar.value = 0
self._progress_bar.max = len(image_files)
if not self._continue_training:
self.apoc.erase_classifier(self._classifier_file.value)
custom_classifier = self._get_training_classifier_instance()
feature_set = self._feature_string.value
channel_index_list = [
self._image_channels.choices.index(channel)
for channel in self._image_channels.value
]
# Iterate over image files, only pulling label files with an identical
# name to the image file. Ensuring that files match by some other
# method would be much more complicated, so I'm leaving it up to the
# user at this point. In addition, the utilities widget saves with
# the same name, so this should be a non-issue, if staying within the
# same workflow.
for idx, image_file in enumerate(image_files):
if not (label_directory / image_file.name).exists():
logger.error('Label file missing for %s', image_file.name)
self._progress_bar.value = idx + 1
continue
logger.info('Training Image %d: %s', idx + 1, image_file.name)
img = AICSImage(image_directory / image_file.name)
channel_img = self._get_channel_image(img, channel_index_list)
lbl = AICSImage(label_directory / image_file.name)
label = lbl.get_image_data('TCZYX', C=0)
# <- this is where setting up dask processing would be useful
try:
custom_classifier.train(
features=feature_set,
image=np.squeeze(channel_img),
ground_truth=np.squeeze(label),
continue_training=True,
)
self._progress_bar.value = idx + 1
except Exception:
logger.exception('Error training %s', image_file)
self._progress_bar.value = idx + 1
continue
self._classifier_statistics_table(custom_classifier)
self._progress_bar.label = f'Trained on {len(image_files)} Images'
logger.removeHandler(handler)
def _get_prediction_classifier_instance(self):
if self._classifier_type.value in self._classifier_type_mapping:
classifier_class = self._classifier_type_mapping[
self._classifier_type.value
]
return classifier_class(
opencl_filename=self._classifier_file.value
)
return None
def batch_predict(self):
from aicsimageio import AICSImage
from aicsimageio.writers import OmeTiffWriter
image_directory, image_files = helpers.get_directory_and_files(
dir_path=self._image_directory.value,
)
log_loc = self._output_directory.value / 'log.txt'
logger, handler = helpers.setup_logger(log_loc)
logger.info(
"""
Classifier: %s
Channels: %s
Num. Files: %d
Image Directory: %s
Output Directory: %s
""",
self._classifier_file.value,
self._image_channels.value,
len(image_files),
image_directory,
self._output_directory.value,
)
# https://github.com/clEsperanto/pyclesperanto_prototype/issues/163
set_wait_for_kernel_finish(True)
self._progress_bar.label = f'Predicting {len(image_files)} Images'
self._progress_bar.value = 0
self._progress_bar.max = len(image_files)
custom_classifier = self._get_prediction_classifier_instance()
channel_index_list = [
self._image_channels.choices.index(channel)
for channel in self._image_channels.value
]
for idx, file in enumerate(image_files):
logger.info('Predicting Image %d: %s', idx + 1, file.name)
img = AICSImage(file)
channel_img = self._get_channel_image(img, channel_index_list)
squeezed_dim_order = helpers.get_squeezed_dim_order(img)
# <- this is where setting up dask processing would be useful
try:
result = custom_classifier.predict(
image=np.squeeze(channel_img)
)
except Exception:
logger.exception('Error predicting %s', file)
self._progress_bar.value = idx + 1
continue
save_data = np.asarray(result)
if save_data.max() > 65535:
save_data = save_data.astype(np.int32)
else:
save_data = save_data.astype(np.int16)
OmeTiffWriter.save(
data=save_data,
uri=self._output_directory.value / (file.stem + '.tiff'),
dim_order=squeezed_dim_order,
channel_names=['Labels'],
physical_pixel_sizes=img.physical_pixel_sizes,
)
del result
self._progress_bar.value = idx + 1
self._progress_bar.label = f'Predicted {len(image_files)} Images'
logger.removeHandler(handler)
def image_train(self):
image_names = [image.name for image in self._image_layers.value]
label_name = self._label_layer.value.name
self._single_result_label.value = (
f'Training on {image_names} using {label_name}'
)
image_list = [image.data for image in self._image_layers.value]
image_stack = np.stack(image_list, axis=0)
label = self._label_layer.value.data
# https://github.com/clEsperanto/pyclesperanto_prototype/issues/163
set_wait_for_kernel_finish(True)
if not self._continue_training:
self.apoc.erase_classifier(self._classifier_file.value)
custom_classifier = self._get_training_classifier_instance()
feature_set = self._feature_string.value
custom_classifier.train(
features=feature_set,
image=np.squeeze(image_stack),
ground_truth=np.squeeze(label),
continue_training=True,
)
self._single_result_label.value = (
f'Trained on {image_names} using {label_name}'
)
def image_predict(self):
set_wait_for_kernel_finish(
True
) # https://github.com/clEsperanto/pyclesperanto_prototype/issues/163
image_names = [image.name for image in self._image_layers.value]
self._single_result_label.value = f'Predicting {image_names}'
image_list = [image.data for image in self._image_layers.value]
image_stack = np.stack(image_list, axis=0)
scale = self._image_layers.value[0].scale
custom_classifier = self._get_prediction_classifier_instance()
result = custom_classifier.predict(image=np.squeeze(image_stack))
# sometimes, input layers may have shape with 1s, like (1,1,10,10)
# however, we are squeezing the input, so the reuslt will have shape
# (10,10), and therefore scale needs to accomodate dropped axes
result_dims = result.ndim
if len(scale) > result_dims:
scale = scale[-result_dims:]
self._viewer.add_labels(result, scale=scale)
self._single_result_label.value = f'Predicted {image_names}'
return result
def insert_custom_feature_string(self):
self._feature_string.value = (
self._custom_apoc_container._feature_string.value
)
return self._feature_string.value
|