-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathcluster.py
1260 lines (1142 loc) · 43.7 KB
/
cluster.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
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
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#
# Copyright 2020-2022 by A. Mathis Group and contributors. All rights reserved.
#
# This project and all its files are licensed under GNU AGPLv3 or later version. A copy is included in https://github.com/AlexEMG/DLC2action/LICENSE.AGPL.
#
import inspect
import os
import pickle
import sys
from collections import defaultdict
from itertools import groupby
from typing import Iterable
import click
import numpy as np
import pyqtgraph as pg
import torch
from PyQt5.Qt import Qt, pyqtSignal
from PyQt5.QtCore import QPoint, QPointF, QRectF, QSize, QTimer
from PyQt5.QtGui import QBrush, QColor, QPen
from PyQt5.QtWidgets import (
QApplication,
QCheckBox,
QComboBox,
QFileDialog,
QFormLayout,
QHBoxLayout,
QLabel,
QLineEdit,
QMessageBox,
QPushButton,
QRadioButton,
QStackedLayout,
QVBoxLayout,
QWidget,
)
from sklearn import datasets, decomposition, manifold
from vispy.scene import SceneCanvas
import annotator
from utils import get_color, get_settings, read_skeleton, read_stack, read_video
from widgets.dialog import (
EpisodeParamsSelector,
EpisodeSelector,
SuggestionParamsSelector,
WarningWindow,
)
from widgets.viewbox import VideoViewBox
class Annotation:
def __init__(self, filename):
self.inds = None
self.cats = None
self.data = None
self.filename = filename
if filename is not None and os.path.exists(filename):
try:
with open(filename, "rb") as f:
_, self.cats, self.inds, self.data = pickle.load(f)
except pickle.UnpicklingError:
pass
def get_len(self):
if self.data is None:
return 0
length = 0
for ind_list in self.data:
for cat_list in ind_list:
if len(cat_list) > 0:
length = max(length, max([x[1] for x in cat_list]))
return length
def generate_array(self):
length = self.get_len()
to_remove = []
for i, x in enumerate(self.inds):
ok = any([len(cat_list) > 0 for cat_list in self.data[i]])
if not ok:
to_remove.append(i)
inds = [x for i, x in enumerate(self.inds) if i not in to_remove]
data = [x for i, x in enumerate(self.data) if i not in to_remove]
for i, x in enumerate(self.cats):
ok = any([len(ind_list[i]) > 0 for ind_list in self.data])
if x.startswith("unknown") or not ok:
to_remove.append(i)
cats = [x for i, x in enumerate(self.cats) if i not in to_remove]
for j in range(len(data)):
data[j] = [x for i, x in enumerate(data[j]) if i not in to_remove]
arr = -100 * np.ones(length * len(inds), len(cats))
for ind_i, ind_list in enumerate(data):
for cat_i, cat_list in enumerate(ind_list):
for start, end, _ in cat_list:
if cats[cat_i].startswith("negative"):
arr[ind_i * length + start, ind_i * length + end, cat_i] = 0
else:
arr[ind_i * length + start, ind_i * length + end, cat_i] = 1
return arr, cats
@staticmethod
def get_filename(filepaths, filenames, video_id, suffix):
for fn, fp in zip(filenames, filepaths):
if fn.split(".")[0] == video_id:
return os.path.join(fp, video_id + suffix + ".pickle")
def unannotated(self):
if self.data is None:
return None
al_points = []
for ind_i, ind in enumerate(self.inds):
labels = np.zeros((len(self.cats), self.get_len()))
for cat in range(len(self.data[ind_i])):
for start, end, _ in self.data[ind_i][cat]:
labels[cat, start:end] = 1
cur = 0
for k, g in groupby(np.sum(labels, axis=0) > 0):
length = len(list(g))
if not k and length > 20:
al_points.append([cur, cur + length, ind])
cur += length
return al_points
def main_labels(self, start, end, ind):
if self.data is None or ind not in self.inds:
return "unknown"
cats = defaultdict(lambda: 0)
for cat, cat_list in zip(self.cats, self.data[self.inds.index(ind)]):
for s, e, _ in cat_list:
if s < end and e > start:
cats[cat] += min(e, end) - max(s, start)
total_annotated = sum(cats.values())
if total_annotated == 0:
return "unknown"
result = set()
for cat, ann in cats.items():
if ann >= (end - start) * 0.1:
if cat.startswith("negative"):
result.add("no behavior")
else:
result.add(cat)
if len(result) > 1:
result = [x for x in result if x != "no behavior"]
if len(result) == 0:
result = ["assorted"]
return " + ".join(sorted(list(result)))
# @staticmethod
# def get_training_data(filepaths, filenames, suffix):
# y = 0
# data = []
# cats = []
# for fp, fn in zip(filepaths, filenames):
# filename = os.path.join(fp, fn.split('.')[0] + suffix + '.pickle')
# arr, cat_list = Annotation(filename).generate_array()
# data.append(arr)
# cats.append(cat_list)
# cats_unique = list(set(cats))
#
# for cat, arr in zip(cats, data):
class PlotWidget(pg.PlotWidget):
clicked = pyqtSignal(QPointF)
dragged = pyqtSignal(list)
released = pyqtSignal()
def __init__(self, select_mode, *args, **kwargs):
super().__init__(*args, **kwargs)
self.pressed = False
self.start = None
self.add = False
self.select_mode = select_mode
def mousePressEvent(self, ev):
if self.select_mode:
modifiers = QApplication.keyboardModifiers()
if modifiers == Qt.ShiftModifier:
self.add = True
else:
self.add = False
p = self.plotItem.vb.mapSceneToView(QPoint(ev.x(), ev.y()))
self.clicked.emit(p)
self.pressed = True
self.start = [p.x(), p.y()]
else:
super().mousePressEvent(ev)
def mouseMoveEvent(self, ev):
if self.select_mode and self.pressed:
p = self.plotItem.vb.mapSceneToView(QPoint(ev.x(), ev.y()))
self.dragged.emit([p.x(), p.y()] + self.start)
else:
super().mouseMoveEvent(ev)
def mouseReleaseEvent(self, ev):
if self.select_mode:
self.pressed = False
self.released.emit()
else:
super().mouseReleaseEvent(ev)
class RectItem(pg.GraphicsObject):
def __init__(self, position, parent=None):
super().__init__(parent)
self.position = position
def paint(self, painter, option, widget=None):
col = QColor("blue")
col.setAlphaF(0.1)
painter.setBrush(col)
pen = QPen(Qt.blue, 0.001)
pen.setStyle(Qt.DotLine)
painter.setPen(pen)
painter.drawRect(QRectF(*self.position))
def boundingRect(self):
return QRectF(*self.position)
class VideoWindow(QWidget):
closed = pyqtSignal(int)
def __init__(self, window, stack, shape, points_df, frames, n, color, label):
super(VideoWindow, self).__init__()
self.layout = QVBoxLayout()
self.setLayout(self.layout)
self.label = QLabel(label)
self.canvas = VideoCanvas(window, stack, shape, points_df, frames, "yellow")
self.layout.addWidget(self.label)
self.layout.addWidget(self.canvas.native)
self.n = n
self.canvas.vb.border_color = color
self.canvas.vb._border_width = 4
self.border_color = 255 * color
self.move(0, 0)
def sizeHint(self):
return QSize(400, 300)
def closeEvent(self, a0):
self.canvas.set_play(False)
self.closed.emit(self.n)
a0.accept()
def color(self):
return list(map(int, self.border_color))
class VideoCanvas(SceneCanvas):
def __init__(self, window, stack, shape, points_df, frames, color):
super(VideoCanvas, self).__init__()
self.unfreeze()
self.window = window
self.play = True
self.speed = 40
_, start, end, ind = frames
self.current = start
self.length = end - start
self.ind = ind
self.skeleton_color = lambda x: color
self.start = start
video = read_stack(stack, start, end, shape)
vb = VideoViewBox(
n_ind=1,
boxes=None,
window=window,
animals=ind,
points_df=points_df,
loaded=[start, end],
current=start,
current_animal=ind,
video=video,
node_transform=self.scene.node_transform,
displayed_animals=[ind],
skeleton_color=self.skeleton_color,
al_mode=False,
al_animal=None,
correct_mode=False,
segmentation=None,
length=self.length,
)
vb.initialize(self.current, 0)
self.central_widget.add_widget(vb)
self.vb = vb
self.timer = QTimer()
self.timer.timeout.connect(self.next)
self.set_play(True)
def on_key_press(self, event):
if event.key.name == "Space":
self.set_play()
elif event.key.name == "Right":
self.next()
elif event.key.name == "Left":
self.prev()
elif event.key.name == "Control":
pass
else:
print(f"canvas didn't recognise key {event.key.name}")
def next(self):
self.set_current_frame(self.current + 1)
def prev(self):
self.set_current_frame(self.current - 1)
def set_play(self, value=None):
if value in [True, False]:
self.play = value
else:
self.play = not self.play
if self.play:
self.timer.start(self.speed)
else:
self.timer.stop()
def set_current_frame(self, value):
next = (value - self.start) % self.length + self.start
self.current = next
self.vb.set_image(self.current, [self.ind], self.skeleton_color, False, None)
self.update()
class LineEdit(QLineEdit):
enter = pyqtSignal()
def keyPressEvent(self, event):
if event.key() == Qt.Key_Enter or event.key() == 16777220:
self.enter.emit()
else:
super(LineEdit, self).keyPressEvent(event)
class ModWindow(QWidget):
def __init__(self, window):
super(ModWindow, self).__init__()
self.window = window
if self.window.method == "raw":
self.parameters = []
self.args = []
else:
self.parameters = inspect.signature(window.dim_red_func).parameters
self.args = list(self.parameters.keys())
for a in ["n_components", "verbose", "n_jobs"]:
if a in self.args:
self.args.remove(a)
self.input_widget = QWidget()
self.input = QFormLayout()
self.input_widget.setLayout(self.input)
self.boxes = []
for par in self.args:
le = LineEdit()
le.enter.connect(self.plot_graph)
self.boxes.append(le)
if par in self.window.args[self.window.method]:
value = str(self.window.args[self.window.method][par])
else:
value = str(self.parameters[par].default)
le.setText(value)
self.input.addRow(par, le)
self.buttons = QHBoxLayout()
self.plot_button = QPushButton("Plot")
self.plot_button.clicked.connect(self.plot_graph)
self.reset_button = QPushButton("Reset to default")
self.reset_button.clicked.connect(self.reset_args)
self.buttons.addWidget(self.reset_button)
self.buttons.addWidget(self.plot_button)
urlLink = f"<a href={self.window.doc_link}>documentation</a>"
self.label1 = QLabel(f"Modify {self.window.method} parameters")
self.label2 = QLabel(f"Here is the {urlLink}")
self.label2.setTextFormat(Qt.RichText)
self.label2.setOpenExternalLinks(True)
self.layout = QVBoxLayout()
self.layout.addWidget(self.label1)
self.layout.addWidget(self.label2)
self.layout.addWidget(self.input_widget)
self.layout.addLayout(self.buttons)
self.setLayout(self.layout)
self.move(0, 0)
def reset_args(self):
self.window.reset_args()
print("reset_args(self):/Args are reset")
self.close()
def stop(self):
print("stop(self)/ Args are reset")
self.close()
def plot_graph(self):
values = [box.text() for box in self.boxes]
corr_values = []
for v in values:
try:
corr_values.append(int(v))
except:
try:
corr_values.append(float(v))
except:
corr_values.append(v)
arg_dict = {par: v for par, v in zip(self.args, corr_values) if v != "None"}
arg_dict["n_components"] = 2
try:
self.window.get_clustering(arg_dict)
self.window.args[self.window.method] = arg_dict
except ValueError as e:
self.value_error(e)
def value_error(self, e):
print(str(e))
msg = QMessageBox()
msg.setWindowTitle("Value error")
msg.setText(str(e) + "\n \n" + "Stop modifying?")
msg.setStandardButtons(QMessageBox.Yes | QMessageBox.No)
msg.buttonClicked.connect(self.error_buttons)
msg.exec_()
def error_buttons(self, button):
if button.text()[1:] == "Yes":
self.stop()
class Console(QWidget):
def __init__(self, window):
super(Console, self).__init__()
self.window = window
self.method_label = QLabel("Dimensionality reduction method:")
self.method_wheel = QComboBox()
self.method_wheel.addItems(self.window.methods)
self.method_wheel.currentTextChanged.connect(self.window.change_method)
self.pars_button = QPushButton("Modify parameters")
self.pars_button.clicked.connect(self.window.modify_pars)
if self.window.method == "raw":
self.pars_button.setEnabled(False)
self.method_checkbox = QCheckBox("Multiple videos open")
self.method_checkbox.toggled.connect(self.window.set_method)
self.close_videos_button = QPushButton("Close all videos")
self.close_videos_button.clicked.connect(self.window.close_all_videos)
self.select_radio = QRadioButton("Select")
self.move_radio = QRadioButton("Move")
self.select_radio.clicked.connect(self.window.set_select_mode)
self.move_radio.clicked.connect(self.window.set_move_mode)
self.move_radio.setChecked(True)
self.save_button = QPushButton("Save selection")
self.save_button.clicked.connect(self.window.save)
self.save_button.setDisabled(True)
self.open_button = QPushButton("Open intervals")
self.open_button.clicked.connect(self.window.open_intervals)
self.open_button.setDisabled(True)
self.auto_button = QPushButton("Autolabel")
self.auto_button.clicked.connect(self.window.autolabel)
self.browse_button = QPushButton("Browse suggestions")
self.browse_button.clicked.connect(self.window.open_suggestions)
self.remain_button = QPushButton("Open unannotated")
self.remain_button.clicked.connect(self.window.get_remainder)
self.radio_layout = QHBoxLayout()
self.radio_layout.addWidget(self.method_checkbox)
self.radio_layout.addWidget(self.close_videos_button)
self.radio_layout.addSpacing(100)
self.radio_layout.addWidget(self.select_radio)
self.radio_layout.addWidget(self.move_radio)
self.radio_layout.addSpacing(100)
self.radio_layout.addWidget(self.save_button)
self.radio_layout.addWidget(self.open_button)
self.radio_layout.addWidget(self.auto_button)
self.radio_layout.addWidget(self.remain_button)
self.radio_layout.addWidget(self.browse_button)
self.method_layout = QHBoxLayout()
self.method_layout.addWidget(self.method_label)
self.method_layout.addWidget(self.method_wheel)
self.method_layout.addWidget(self.pars_button)
self.layout = QVBoxLayout()
self.layout.addLayout(self.method_layout)
self.layout.addLayout(self.radio_layout)
self.setLayout(self.layout)
class MainWindow(QWidget):
switch = pyqtSignal()
status = pyqtSignal(str)
def __init__(
self,
filenames,
filepaths,
open_settings,
config_file,
features,
skeleton_folder,
annotation_folder,
feature_folder,
feature_suffix,
dlc2action_name,
dlc2action_path,
skip_dlc2action,
*args,
**kwargs,
):
super(MainWindow, self).__init__(*args, **kwargs)
if annotation_folder is None:
annotation_folder = feature_folder
self.settings_file = config_file
self.settings = get_settings(self.settings_file, open_settings)
self.filepaths = filepaths
self.filenames = filenames
self.feature_files = features
self.annotation_files = []
self.annotation_folder = annotation_folder
self.feature_suffix = feature_suffix
self.feature_folder = feature_folder
if annotation_folder is not None and not os.path.exists(annotation_folder):
os.mkdir(annotation_folder)
for fn, fp in zip(self.filenames, self.filepaths):
ann_folder = annotation_folder if annotation_folder is not None else fp
ann_path = os.path.join(
ann_folder, fn.split(".")[0] + self.settings["suffix"]
)
self.annotation_files.append(ann_path)
self.skeleton_files = []
for fn, fp in zip(self.filenames, self.filepaths):
sk_folder = skeleton_folder if skeleton_folder is not None else fp
sk_path = None
for s in self.settings["DLC_suffix"]:
path = os.path.join(sk_folder, fn.split(".")[0] + s)
if os.path.exists(path):
sk_path = path
break
self.skeleton_files.append(sk_path)
self.parameters = [
filenames,
filepaths,
False,
config_file,
features,
skeleton_folder,
annotation_folder,
feature_folder,
feature_suffix,
dlc2action_name,
dlc2action_path,
skip_dlc2action,
]
cwd = os.getcwd()
try:
with open("colors.txt") as f:
self.colors = (
np.array([list(map(int, line.split())) for line in f.readlines()]) / 255
)
except:
if not cwd.endswith('/Project_Config'):
os.chdir(os.path.join(os.getcwd(),'Project_Config'))
with open("colors.txt") as f:
self.colors = (
np.array([list(map(int, line.split())) for line in f.readlines()]) / 255
)
os.chdir(cwd)
self.video_mode = "same"
self.open_videos = {}
self.color_i = 0
self.setWindowTitle("Clustering")
self.method = "raw"
self.methods = ["raw", "t-SNE", "PCA", "LLE", "ICA", "FA"]
self.rect_position = [0, 0, 0, 0]
self.chosen = []
self.visited = set()
self.select_mode = False
self.load_files()
self.graph = PlotWidget(self.select_mode)
self.graph.setBackground("w")
self.graph.dragged.connect(self.set_rect)
self.graph.released.connect(self.on_release)
self.args = {x: {"n_components": 2} for x in self.methods}
self.get_data() # TODO: self.X, self.y (load from .npy), self.frames (al_points style)
self.get_clustering(self.args[self.method])
# self.stack, self.shape, self.len = read_video('/Users/liza/data/CAM2_2021_06_25_083528.mp4')
# short_len = (self.len - 100) // 100
# self.frames = [(i, i+100) for i in range(0, self.len, short_len)][:100]
# self.skeleton_file = None
self.dlc2action_name = dlc2action_name
self.dlc2action_path = dlc2action_path
self.initialize_dlc2action_project(skip_dlc2action)
self.console = Console(self)
self.video_layout = QHBoxLayout()
self.video_layout.addWidget(self.console)
self.main_layout = QStackedLayout()
self.main_layout.setStackingMode(QStackedLayout.StackAll)
self.main_layout.addWidget(self.graph)
self.layout = QVBoxLayout()
self.layout.addLayout(self.video_layout)
self.layout.addLayout(self.main_layout)
self.setLayout(self.layout)
def get_color(self, name):
return get_color(self.colors, name)
def initialize_dlc2action_project(self, skip):
if skip:
self.dlc2action_name = None
return
from dlc2action.project import Project
if self.dlc2action_name is None:
i = 0
while not Project.project_name_available(
self.dlc2action_path, f"cluster_project_{i}"
):
i += 1
self.dlc2action_name = f"cluster_project_{i}"
project = Project(
self.dlc2action_name,
data_type="features",
annotation_type="dlc",
projects_path=self.dlc2action_path,
data_path=self.feature_folder,
annotation_path=self.annotation_folder,
)
project.update_parameters(
{
"data": {
"annotation_suffix": self.settings["suffix"],
"feature_suffix": self.feature_suffix,
"filter_annotated": True,
"filter_background": True,
"behaviors": None,
"transpose_features": True,
},
"general": {
"metric_functions": {"f1"},
"only_load_annotated": True,
"len_segment": 128,
"overlap": 1 / 2,
"model_name": "mlp",
"exclusive": False,
},
"training": {
"test_frac": 0,
"val_frac": 0,
"to_ram": False,
"num_epochs": 100,
"lr": 1e-4,
"device": "cuda" if torch.cuda.is_available() else "cpu",
"partition_method": "random",
},
"metrics": {
"f1": {"average": "none"},
},
}
)
def load_animals(self, skeleton_file):
if skeleton_file is not None:
try:
points_df, _ = read_skeleton(
skeleton_file, data_type=self.settings["data_type"]
)
except:
print("skeleton file is invalid or does not exist")
points_df = None
else:
points_df = None
return points_df
def load_skeleton(self, skeleton_file):
points_df = self.load_animals(skeleton_file)
return points_df
def load_files(self):
self.loading_dict = {
fn.split(".")[0]: (fp, fn, skeleton)
for fp, fn, skeleton in zip(
self.filepaths, self.filenames, self.skeleton_files
)
}
self.points_df_dict = {}
self.stacks = {}
self.shapes = {}
def open_video(self, name):
fp, fn, skeleton = self.loading_dict[name]
name = fn.split(".")[0]
stack, shape, _ = read_video(os.path.join(fp, fn))
self.stacks[name] = stack
self.shapes[name] = shape
self.points_df_dict[name] = self.load_skeleton(skeleton)
def video_closed(self, n):
self.open_videos.pop(n)
self.plot_clusters()
def clicked(self, point, ev):
data_list = list(self.scatter.data)
points_list = [x[9] for x in data_list]
n = points_list.index(ev[0])
self.visited.update([n])
if self.video_mode == "same":
self.close_all_videos()
vid, start, end, ind = self.frames[n]
label = self.labels[n]
if label in self.label_dict:
label = self.label_dict[label]
if vid not in self.stacks:
self.open_video(vid)
if self.points_df_dict[vid] is not None:
points_df = self.points_df_dict[vid].get_range(start, end, ind)
else:
points_df = None
vw = VideoWindow(
self,
self.stacks[vid],
self.shapes[vid],
points_df,
self.frames[n],
n,
self.colors[self.color_i],
label,
)
vw.closed.connect(self.video_closed)
self.color_i = (self.color_i + 1) % len(self.colors)
vw.show()
self.open_videos[n] = vw
self.plot_clusters()
def modify_pars(self):
self.mod = ModWindow(self)
self.mod.show()
def get_data(self):
if sum(x is not None for x in self.feature_files) == 0:
import random
n = 100
X, y = datasets.make_blobs(
n_samples=n, cluster_std=[1.0, 2.5, 0.5], n_features=5
)
self.label_dict = {0: "zero", 1: "one", 2: "two"}
self.data = X
self.labels = []
videos = [f.split(".")[0] for f in self.filenames]
videos_rand = [videos[random.randint(0, len(videos) - 1)] for _ in range(n)]
start_frames = list(range(0, 1000, 50))
self.frames = [
(
v,
start_frames[i % len(start_frames)],
start_frames[i % len(start_frames)] + 50,
"ind0",
)
for i, v in enumerate(videos_rand)
]
for video in videos:
annotation = Annotation(
Annotation.get_filename(
self.filepaths, self.filenames, video, self.settings["suffix"]
)
)
for v, s, e, clip in self.frames:
if v != video:
continue
main_labels = annotation.main_labels(s, e, clip)
self.labels.append(main_labels)
labels = list(set(self.labels))
self.label_dict = {i: x for i, x in enumerate(labels)}
inv_dict = {x: i for i, x in enumerate(labels)}
self.labels = [inv_dict[x] for x in self.labels]
else:
self.filenames = [
x
for i, x in enumerate(self.filenames)
if self.feature_files[i] is not None
]
self.filepaths = [
x
for i, x in enumerate(self.filepaths)
if self.feature_files[i] is not None
]
self.feature_files = [x for x in self.feature_files if x is not None]
self.frames = []
self.data = []
self.labels = []
frames_step = 256
for file_i, file in enumerate(self.feature_files):
if file.endswith("npy"):
video_dict = np.load(file, allow_pickle=True).item()
else:
with open(file, "rb") as f:
video_dict = pickle.load(f)
video = self.filenames[file_i]
annotation = Annotation(self.annotation_files[file_i])
min_frames = video_dict.pop("min_frames")
clips = list(video_dict.keys())
for clip in clips:
if clip in ["max_frames", "video_tag"]:
continue
clip_arr = video_dict.pop(clip)
# print(f'{clip_arr.shape=}')
n = clip_arr.shape[0]
if n != 2048:
# if not ((n & (n-1) == 0) and n != 0):
clip_arr = clip_arr.T
for s in range(0, clip_arr.shape[-1], frames_step):
end = min(clip_arr.shape[-1], s + frames_step)
main_labels = annotation.main_labels(
s + min_frames[clip], end + min_frames[clip], clip
)
self.frames.append(
(
video.split(".")[0],
s + min_frames[clip],
end + min_frames[clip],
clip,
)
)
self.data.append(clip_arr[:, s:end].mean(-1))
self.labels.append(main_labels)
labels = sorted(list(set(self.labels)))
self.label_dict = {i: x for i, x in enumerate(labels)}
inv_dict = {x: i for i, x in enumerate(labels)}
self.labels = [inv_dict[x] for x in self.labels]
if isinstance(self.data[0], np.ndarray):
self.data = np.stack(self.data, 0)
else:
self.data = torch.stack(self.data, 0).numpy()
def reset_args(self):
print("reset_args(self): Window /Args are reset")
self.args[self.method] = {"n_components": 2}
self.get_clustering(self.args[self.method])
def get_clustering(self, args):
method_dict = {
"PCA": decomposition.PCA,
"t-SNE": manifold.TSNE,
"LLE": manifold.LocallyLinearEmbedding,
"FA": decomposition.FactorAnalysis,
"ICA": decomposition.FastICA,
}
link_dict = {
"PCA": "https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html",
"t-SNE": "https://scikit-learn.org/stable/modules/generated/sklearn.manifold.TSNE.html",
"LLE": "https://scikit-learn.org/stable/modules/generated/sklearn.manifold.LocallyLinearEmbedding.html",
"FA": "https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.FactorAnalysis.html",
"ICA": "https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.FastICA.html#sklearn.decomposition.FastICA",
}
if self.method == "raw":
self.components = self.data[:, :2]
self.doc_link = ""
else:
self.dim_red_func = method_dict[self.method]
self.doc_link = link_dict[self.method]
self.dim_red = self.dim_red_func(**args)
self.components = self.dim_red.fit_transform(self.data)
self.plot_clusters()
def make_brush(self, x, n):
if x == -100 or self.label_dict[x] == "unknown":
col = QColor("gray")
col.setAlphaF(0.2)
elif self.label_dict[x] == "no behavior":
col = pg.mkColor(110, 110, 110)
else:
col = pg.mkColor(self.get_color(self.label_dict[x]) * 255)
if n in self.chosen and x != -100:
if col == Qt.blue:
col.setAlphaF(0.1)
else:
col.setAlphaF(0.7)
return QBrush(col)
def make_pen(self, n):
if n in self.chosen:
return QPen(Qt.blue, 0.3)
else:
if n in self.visited:
col = pg.mkColor(121, 92, 178)
return QPen(col, 0.2)
else:
return QPen(Qt.gray, 0.2)
def plot_clusters(self):
self.graph.clear()
brush = [self.make_brush(int(x), n) for n, x in enumerate(self.labels)]
pen = [self.make_pen(n) for n in range(len(self.components))]
self.scatter = pg.ScatterPlotItem(
pos=self.components,
hoverable=True,
hoverBrush=QBrush(Qt.gray),
brush=brush,
symbol="o",
pen=pen,
)
brush = []
for n in self.open_videos:
brush.append(QBrush(QColor(*self.open_videos[n].color())))
self.scatter.addPoints(
pos=self.components[list(self.open_videos.keys())],
hoverable=False,
hoverBrush=QBrush(Qt.gray),
brush=brush,
symbol="x",
size=23,
pen=QPen(Qt.white, 0.1),
)
self.graph.addItem(self.scatter)
self.rect = RectItem(self.rect_position)
self.graph.addItem(self.rect)
self.scatter.sigClicked.connect(self.clicked)
self.update()
def change_method(self, method):
self.method = method
if self.method == "raw":
self.console.pars_button.setEnabled(False)
else:
self.console.pars_button.setEnabled(True)
self.reset_args()
self.get_clustering(self.args[self.method])
def set_method(self, ev):
if not ev:
self.video_mode = "same"
self.open_videos = {}
else:
self.video_mode = "new"
def close_all_videos(self):
print("close_all_videos(self):")
for n in list(self.open_videos.keys()):
self.open_videos[n].close()
def set_rect(self, event):
cur_x, cur_y, start_x, start_y = event
self.rect_position = [start_x, start_y, cur_x - start_x, cur_y - start_y]
self.plot_clusters()
def on_release(self):
s_x, s_y, w, h = self.rect_position
e_x = s_x + w
e_y = s_y + h
x = (self.components[:, 0] > min(s_x, e_x)) & (
self.components[:, 0] < max(s_x, e_x)
)
y = (self.components[:, 1] > min(s_y, e_y)) & (
self.components[:, 1] < max(s_y, e_y)
)
if self.graph.add:
self.chosen += list(np.where(x & y)[0])
else:
self.chosen = list(np.where(x & y)[0])
self.rect_position = [0, 0, 0, 0]
self.plot_clusters()
self.console.save_button.setDisabled(len(self.chosen) == 0)
self.console.open_button.setDisabled(len(self.chosen) == 0)
def set_select_mode(self):
self.select_mode = True
self.graph.select_mode = True
def set_move_mode(self):
self.select_mode = False
self.graph.select_mode = False
def save(self):
self.output_file = QFileDialog.getSaveFileName(self, "Save file")[0]
with open(self.output_file, "wb") as f:
pickle.dump(self.chosen, f)
def reopen(self):
print("reopen(self): window closed")
self.get_data()
self.show()
def get_remainder(self):