forked from mpkuse/cartwheel_train
-
Notifications
You must be signed in to change notification settings - Fork 0
/
PandaRender.py
1378 lines (1009 loc) · 45.6 KB
/
PandaRender.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
""" The training-render file. (class TrainRenderer)
Update: This class is render only class. It does not take the mainloop()
control. Basically need to call function step().
Defines a rendering class. Defines a spinTask (panda3d) which basicalyl
renders 16-cameras at a time and sets them into a CPU-queue. This queue
is emtied by calls to step(). May build more purpose-build steps()s
update: There are 3 Panda renderer in this file, viz. TrainRenderer,
TestRenderer, NetVLADRenderer. For comments on what each does check out
those classes. For general usage of the renderers see `test_render.py`
"""
# Panda3d
from direct.showbase.ShowBase import ShowBase
from panda3d.core import *
from direct.stdpy.threading2 import Thread
# Usual Math and Image Processing
import numpy as np
import cv2
from scipy import interpolate
# import caffe
# import tensorflow as tf
# Other System Libs
import os
import argparse
import Queue
import copy
import time
import code
import pickle
# Custom-Misc
import TerminalColors
import CubeMaker
import PathMaker
class TrainRenderer(ShowBase):
renderIndx=0
# Basic Mesh & Camera Setup
def loadAllTextures(self, mesh, basePath, silent=True):
""" Loads texture files for a mesh """
c = 0
for child in mesh.getChildren():
submesh_name = child.get_name()
submesh_texture = basePath + submesh_name[:-5] + 'tex0.jpg'
child.setTexture( self.loader.loadTexture(submesh_texture) )
if silent == False:
print 'Loading texture file : ', submesh_texture
c = c + 1
print self.tcolor.OKGREEN, "Loaded ", c, "textures", self.tcolor.ENDC
def setupMesh(self):
""" Loads the .obj files. Will load mesh sub-divisions separately """
print 'Attempt Loading Mesh VErtices, FAces'
self.cyt = self.loader.loadModel( 'model_l/l6/level_6_0_0.obj' )
self.cyt2 = self.loader.loadModel( 'model_l/l6/level_6_128_0.obj' )
self.low_res = self.loader.loadModel( 'model_l/l3/level_3_0_0.obj' )
print self.tcolor.OKGREEN, 'Done Loading Vertices', self.tcolor.ENDC
print 'Attempt Loading Textures'
self.loadAllTextures( self.cyt, 'model_l/l6/')
self.loadAllTextures( self.cyt2, 'model_l/l6/')
self.loadAllTextures( self.low_res, 'model_l/l3/')
print self.tcolor.OKGREEN, 'Done Loading Textures', self.tcolor.ENDC
def positionMesh(self):
""" WIll have to manually adjust this for ur mesh. I position the
center where I fly my drone and oriented in ENU (East-north-up)
cords for easy alignment of GPS and my cordinates. If your model
is not metric scale will have to adjust for that too"""
self.cyt.setPos( 140,-450, 150 )
self.cyt2.setPos( 140,-450, 150 )
self.low_res.setPos( 140,-450, 150 )
self.cyt.setHpr( 198, -90, 0 )
self.cyt2.setHpr( 198, -90, 0 )
self.low_res.setHpr( 198, -90, 0 )
self.cyt.setScale(172)
self.cyt2.setScale(172)
self.low_res.setScale(172)
def customCamera(self, nameIndx):
lens = self.camLens
lens.setFov(83)
print 'self.customCamera : Set FOV at 83'
my_cam = Camera("cam"+nameIndx, lens)
my_camera = self.scene0.attachNewNode(my_cam)
# my_camera = self.render.attachNewNode(my_cam)
my_camera.setName("camera"+nameIndx)
return my_camera
def customDisplayRegion(self, rows, cols):
rSize = 1.0 / rows
cSize = 1.0 / cols
dr_list = []
for i in range(0,rows):
for j in range(0,cols):
# print i*rSize, (i+1)*rSize, j*cSize, (j+1)*cSize
dr_i = self.win2.makeDisplayRegion(i*rSize, (i+1)*rSize, j*cSize, (j+1)*cSize)
dr_i.setSort(-5)
dr_list.append( dr_i )
return dr_list
## Gives a random 6-dof pose. Need to set params manually here.
## X,Y,Z, Yaw(abt Z-axis), Pitch(abt X-axis), Roll(abt Y-axis)
## @param No : no inputs
def monte_carlo_sample(self):
# mc_X_min etc are set in constructor
X = np.random.uniform(self.mc_X_min,self.mc_X_max)
Y = np.random.uniform(self.mc_Y_min,self.mc_Y_max)
Z = np.random.uniform(self.mc_Z_min,self.mc_Z_max)
yaw = np.random.uniform( self.mc_yaw_min, self.mc_yaw_max)
roll = 0#np.random.uniform( self.mc_roll_min, self.mc_roll_max)
pitch = 0#np.random.uniform( self.mc_pitch_min, self.mc_pitch_max)
return X,Y,Z, yaw,roll,pitch
## Annotation-helpers for self.render
def putBoxes(self,X,Y,Z,r=1.,g=0.,b=0., scale=1.0):
cube_x = CubeMaker.CubeMaker().generate()
cube_x.setColor(r,g,b)
cube_x.setScale(scale)
cube_x.reparentTo(self.render)
cube_x.setPos(X,Y,Z)
## Set a cube in 3d env
def putTrainingBox(self,task):
cube = CubeMaker.CubeMaker().generate()
cube.setTransparency(TransparencyAttrib.MAlpha)
cube.setAlphaScale(0.5)
# cube.setScale(10)
# mc_X_min etc are set in constructor
sx = 0.5 * (self.mc_X_max - self.mc_X_min)
sy = 0.5 * (self.mc_Y_max - self.mc_Y_min)
sz = 0.5 * (self.mc_Z_max - self.mc_Z_min)
ax = 0.5 * (self.mc_X_max + self.mc_X_min)
ay = 0.5 * (self.mc_Y_max + self.mc_Y_min)
az = 0.5 * (self.mc_Z_max + self.mc_Z_min)
cube.setSx(sx)
cube.setSy(sy)
cube.setSz(sz)
cube.reparentTo(self.render)
cube.setPos(ax,ay,az)
## Task. This task draw the XYZ axis
def putAxesTask(self,task):
if (task.frame / 10) % 2 == 0:
cube_x = CubeMaker.CubeMaker().generate()
cube_x.setColor(1.0,0.0,0.0)
cube_x.setScale(1)
cube_x.reparentTo(self.render)
cube_x.setPos(task.frame,0,0)
cube_y = CubeMaker.CubeMaker().generate()
cube_y.setColor(0.0,1.0,0.0)
cube_y.setScale(1)
cube_y.reparentTo(self.render)
cube_y.setPos(0,task.frame,0)
cube_z = CubeMaker.CubeMaker().generate()
cube_z.setColor(0.0,0.0,1.0)
cube_z.setScale(1)
cube_z.reparentTo(self.render)
cube_z.setPos(0,0,task.frame)
if task.time > 25:
return None
return task.cont
## Render-n-Learn task
##
## set pose in each camera <br/>
## make note of the poses just set as this will take effect next <br/>
## Retrive Rendered Data <br/>
## Cut rendered data into individual image. Note rendered data will be 4X4 grid of images <br/>
## Put imX into the queue <br/>
def renderNlearnTask(self, task):
if task.time < 2: #do not do anything for 1st 2 sec
return task.cont
# print randX, randY, randZ
#
## set pose in each camera
# Note: The texture is grided images in a col-major format
poses = np.zeros( (len(self.cameraList), 4), dtype='float32' )
for i in range(len(self.cameraList)):
randX,randY, randZ, randYaw, randPitch, randRoll = self.monte_carlo_sample()
# if i<4 :
# randX = (i) * 30
# else:
# randX = 0
#
# randY = 0#task.frame
# randZ = 80
# randYaw = 0
# randPitch = 0
# randRoll = 0
self.cameraList[i].setPos(randX,randY,randZ)
self.cameraList[i].setHpr(randYaw,-90+randPitch,0+randRoll)
poses[i,0] = randX
poses[i,1] = randY
poses[i,2] = randZ
poses[i,3] = randYaw
# self.putBoxes(randX,randY,randZ, scale=0.3)
#
# if task.frame < 100:
# return task.cont
# else:
# return None
## make note of the poses just set as this will take effect next
if TrainRenderer.renderIndx == 0:
TrainRenderer.renderIndx = TrainRenderer.renderIndx + 1
self.prevPoses = poses
return task.cont
#
## Retrive Rendered Data
tex = self.win2.getScreenshot()
A = np.array(tex.getRamImageAs("RGB")).reshape(960,1280,3)
# A = np.zeros((960,1280,3))
# A_bgr = cv2.cvtColor(A.astype('uint8'),cv2.COLOR_RGB2BGR)
# cv2.imwrite( str(TrainRenderer.renderIndx)+'.png', A_bgr.astype('uint8') )
# myTexture = self.win2.getTexture()
# print myTexture
# retrive poses from prev render
texPoses = self.prevPoses
#
## Cut rendered data into individual image. Note rendered data will be 4X4 grid of images
#960 rows and 1280 cols (4x4 image-grid)
nRows = 240
nCols = 320
# Iterate over the rendered texture in a col-major format
c=0
if self.q_imStack.qsize() < 150:
for j in range(4): #j is for cols-indx
for i in range(4): #i is for rows-indx
#print i*nRows, j*nCols, (i+1)*nRows, (j+1)*nCols
im = A[i*nRows:(i+1)*nRows,j*nCols:(j+1)*nCols,:]
#imX = im.astype('float32')/255. - .5 # TODO: have a mean image
#imX = (im.astype('float32') - 128.0) /128.
imX = im.astype('float32') #- self.meanImage
## Put imX into the queue
# do not queue up if queue size begin to exceed 150
self.q_imStack.put( imX )
self.q_labelStack.put( texPoses[c,:] )
# fname = '__'+str(poses[c,0]) + '_' + str(poses[c,1]) + '_' + str(poses[c,2]) + '_' + str(poses[c,3]) + '_'
# cv2.imwrite( str(TrainRenderer.renderIndx)+'__'+str(i)+str(j)+fname+'.png', imX.astype('uint8') )
c = c + 1
else:
if self.queue_warning:
print 'q_imStack.qsize() > 150. Queue is filled, not retriving the rendered data'
#
# Call caffe iteration (reads from q_imStack and q_labelStack)
# Possibly upgrade to TensorFlow
# self.learning_iteration()
# if( TrainRenderer.renderIndx > 50 ):
# return None
#
# Prep for Next Iteration
TrainRenderer.renderIndx = TrainRenderer.renderIndx + 1
self.prevPoses = poses
return task.cont
## Execute 1-step.
##
## This function is to be called from outside to render once. This is a wrapper for app.taskMgr.step()
def step(self, batchsize):
""" One rendering.
This function needs to be called from outside in a loop for continous rendering
Returns 2 variables. One im_batch and another label
"""
# ltimes = int( batchsize/16 ) + 1
# print 'Render ', ltimes, 'times'
# for x in range(ltimes):
# Note: 2 renders sometime fails. Donno exactly what happens :'(
# Instead I do app.taskMgr.step() in the main() instead, once and 1 time here. This seem to work OK
# self.taskMgr.step()
# Thread.sleep(0.1)
if self.q_imStack.qsize() < 16*5:
self.taskMgr.step()
# print 'Queues Status (imStack=%d,labelStack=%d)' %(self.q_imStack.qsize(), self.q_labelStack.qsize())
# TODO: Check validity of batchsize. Also avoid hard coding the thresh for not retriving from queue.
im_batch = np.zeros((batchsize,240,320,3))
label_batch = np.zeros((batchsize,4))
# assert self.q_imStack.qsize() > 16*5
if self.q_imStack.qsize() >= 16*5:
# get a batch out
for i in range(batchsize):
im = self.q_imStack.get() #240x320x3 RGB
y = self.q_labelStack.get()
# print 'retrive', i
#remember to z-normalize
im_batch[i,:,:,0] = copy.deepcopy(im[:,:,0])#self.zNormalized( copy.deepcopy(im[:,:,0]) )
im_batch[i,:,:,1] = copy.deepcopy(im[:,:,1])#self.zNormalized( copy.deepcopy(im[:,:,1]) )
im_batch[i,:,:,2] = copy.deepcopy(im[:,:,2])#self.zNormalized( copy.deepcopy(im[:,:,2]) )
label_batch[i,0] = copy.deepcopy( y[0] )
label_batch[i,1] = copy.deepcopy( y[1] )
label_batch[i,2] = copy.deepcopy( y[2] )
label_batch[i,3] = copy.deepcopy( y[3] )
else:
return None, None
f_im = 'im_batch.pickle'
f_lab = 'label_batch.pickle'
print 'Loading : ', f_im, f_lab
with open( f_im, 'rb' ) as handle:
im_batch = pickle.load(handle )
with open( f_lab, 'rb' ) as handle:
label_batch = pickle.load(handle )
print 'Done.@!'
# im_batch = copy.deepcopy( self.X_im_batch )
# # label_batch = copy.deepcopy( self.X_label_batch )
#
r0 = np.random.randint( 0, im_batch.shape[0], batchsize )
# r1 = np.random.randint( 0, im_batch.shape[0], batchsize )
im_batch = im_batch[r0]
label_batch = label_batch[r0]
# Note:
# What is being done here is a bit of a hack. The thing is
# in the mainloop() ie. in train_tf_decop.py doesn't allow any
# if statements. So, I have instead saved a few example-renders on a
# pickle-file. If the queue is not sufficiently filled i just return
# from the saved file.
return im_batch, label_batch
def __init__(self, queue_warning=True):
ShowBase.__init__(self)
self.taskMgr.add( self.renderNlearnTask, "renderNlearnTask" ) #changing camera poses
self.taskMgr.add( self.putAxesTask, "putAxesTask" ) #draw co-ordinate axis
self.taskMgr.add( self.putTrainingBox, "putTrainingBox" )
self.queue_warning = queue_warning #suppress the warning of queue full if this var is True
# Set up training area. This is used in monte_carlo_sample() and putTrainingBox()
self.mc_X_max = 300
self.mc_X_min = -300
self.mc_Y_max = 360
self.mc_Y_min = -360
self.mc_Z_max = 120
self.mc_Z_min = 45
self.mc_yaw_max = 60
self.mc_yaw_min = -60
self.mc_roll_max = 5
self.mc_roll_min = -5
self.mc_pitch_max = 5
self.mc_pitch_min = -5
# # Note params
# self.PARAM_TENSORBOARD_PREFIX = TENSORBOARD_PREFIX
# self.PARAM_MODEL_SAVE_PREFIX = MODEL_SAVE_PREFIX
# self.PARAM_MODEL_RESTORE = MODEL_RESTORE
#
# self.PARAM_WRITE_SUMMARY_EVERY = WRITE_SUMMARY_EVERY
# self.PARAM_WRITE_TF_MODEL_EVERY = WRITE_TF_MODEL_EVERY
# Misc Setup
self.render.setAntialias(AntialiasAttrib.MAuto)
self.setFrameRateMeter(True)
self.tcolor = TerminalColors.bcolors()
#
# Set up Mesh (including load, position, orient, scale)
self.setupMesh()
self.positionMesh()
# Custom Render
# Important Note: self.render displays the low_res and self.scene0 is the images to retrive
self.scene0 = NodePath("scene0")
# cytX = copy.deepcopy( cyt )
self.low_res.reparentTo(self.render)
self.cyt.reparentTo(self.scene0)
self.cyt2.reparentTo(self.scene0)
#
# Make Buffering Window
bufferProp = FrameBufferProperties().getDefault()
props = WindowProperties()
props.setSize(1280, 960)
win2 = self.graphicsEngine.makeOutput( pipe=self.pipe, name='wine1',
sort=-1, fb_prop=bufferProp , win_prop=props,
flags=GraphicsPipe.BFRequireWindow)
#flags=GraphicsPipe.BFRefuseWindow)
# self.window = win2#self.win #dr.getWindow()
self.win2 = win2
# self.win2.setupCopyTexture()
# Adopted from : https://www.panda3d.org/forums/viewtopic.php?t=3880
#
# Set Multiple Cameras
self.cameraList = []
for i in range(4*4):
print 'Create camera#', i
self.cameraList.append( self.customCamera( str(i) ) )
# Disable default camera
# dr = self.camNode.getDisplayRegion(0)
# dr.setActive(0)
#
# Set Display Regions (4x4)
dr_list = self.customDisplayRegion(4,4)
#
# Setup each camera
for i in range(len(dr_list)):
dr_list[i].setCamera( self.cameraList[i] )
#
# Set buffered Queues (to hold rendered images and their positions)
# each queue element will be an RGB image of size 240x320x3
self.q_imStack = Queue.Queue()
self.q_labelStack = Queue.Queue()
print self.tcolor.OKGREEN, '\n##########\n'+'Panda3d Renderer Initialization Complete'+'\n##########\n', self.tcolor.ENDC
class TestRenderer(ShowBase):
renderIndx=0
## Basic Mesh & Camera Setup
def loadAllTextures(self, mesh, basePath, silent=True):
""" Loads texture files for a mesh """
c = 0
for child in mesh.getChildren():
submesh_name = child.get_name()
submesh_texture = basePath + submesh_name[:-5] + 'tex0.jpg'
child.setTexture( self.loader.loadTexture(submesh_texture) )
if silent == False:
print 'Loading texture file : ', submesh_texture
c = c + 1
print self.tcolor.OKGREEN, "Loaded ", c, "textures", self.tcolor.ENDC
def setupMesh(self):
""" Loads the .obj files. Will load mesh sub-divisions separately """
print 'Attempt Loading Mesh VErtices, FAces'
self.cyt = self.loader.loadModel( 'model_l/l6/level_6_0_0.obj' )
self.cyt2 = self.loader.loadModel( 'model_l/l6/level_6_128_0.obj' )
self.low_res = self.loader.loadModel( 'model_l/l0/level_0_0_0.obj' )
print self.tcolor.OKGREEN, 'Done Loading Vertices', self.tcolor.ENDC
print 'Attempt Loading Textures'
self.loadAllTextures( self.cyt, 'model_l/l6/')
self.loadAllTextures( self.cyt2, 'model_l/l6/')
self.loadAllTextures( self.low_res, 'model_l/l0/')
print self.tcolor.OKGREEN, 'Done Loading Textures', self.tcolor.ENDC
def positionMesh(self):
""" WIll have to manually adjust this for ur mesh. I position the
center where I fly my drone and oriented in ENU (East-north-up)
cords for easy alignment of GPS and my cordinates. If your model
is not metric scale will have to adjust for that too"""
self.cyt.setPos( 140,-450, 150 )
self.cyt2.setPos( 140,-450, 150 )
self.low_res.setPos( 140,-450, 150 )
self.cyt.setHpr( 198, -90, 0 )
self.cyt2.setHpr( 198, -90, 0 )
self.low_res.setHpr( 198, -90, 0 )
self.cyt.setScale(172)
self.cyt2.setScale(172)
self.low_res.setScale(172)
def customCamera(self, nameIndx):
lens = self.camLens
lens.setFov(83)
print 'self.customCamera : Set FOV at 83'
my_cam = Camera("cam"+nameIndx, lens)
my_camera = self.scene0.attachNewNode(my_cam)
# my_camera = self.render.attachNewNode(my_cam)
my_camera.setName("camera"+nameIndx)
return my_camera
def customDisplayRegion(self, rows, cols):
rSize = 1.0 / rows
cSize = 1.0 / cols
dr_list = []
for i in range(0,rows):
for j in range(0,cols):
# print i*rSize, (i+1)*rSize, j*cSize, (j+1)*cSize
dr_i = self.win2.makeDisplayRegion(i*rSize, (i+1)*rSize, j*cSize, (j+1)*cSize)
dr_i.setSort(-5)
dr_list.append( dr_i )
return dr_list
def monte_carlo_sample(self):
""" Gives a random 6-dof pose. Need to set params manually here.
X,Y,Z, Yaw(abt Z-axis), Pitch(abt X-axis), Roll(abt Y-axis) """
X = np.random.uniform(-50,50)
Y = np.random.uniform(-100,100)
Z = np.random.uniform(50,100)
yaw = np.random.uniform(-60,60)
roll = np.random.uniform(-5,5)
pitch = np.random.uniform(-5,5)
return X,Y,Z, yaw,roll,pitch
## Annotation-helpers for self.render
def putBoxes(self,X,Y,Z,r=1.,g=0.,b=0., scale=1.0):
cube_x = CubeMaker.CubeMaker().generate()
cube_x.setColor(r,g,b)
cube_x.setScale(scale)
cube_x.reparentTo(self.render)
cube_x.setPos(X,Y,Z)
def putAxesTask(self,task):
cube_x = CubeMaker.CubeMaker().generate()
cube_x.setColor(1.0,0.0,0.0)
cube_x.setScale(1)
cube_x.reparentTo(self.render)
cube_x.setPos(task.frame,0,0)
cube_y = CubeMaker.CubeMaker().generate()
cube_y.setColor(0.0,1.0,0.0)
cube_y.setScale(1)
cube_y.reparentTo(self.render)
cube_y.setPos(0,task.frame,0)
cube_z = CubeMaker.CubeMaker().generate()
cube_z.setColor(0.0,0.0,1.0)
cube_z.setScale(1)
cube_z.reparentTo(self.render)
cube_z.setPos(0,0,task.frame)
if task.time > 25:
return None
return task.cont
## Render-n-Learn task.
## Sets the camera position (1-cam only) with spline or your choice (see init).
### Renders the image at that point and queue the image and its position (pose)
def renderNtestTask(self, task):
if task.frame < 50: #do not do anything for 50 ticks, as spline's 1st node is at t=50
return task.cont
# print randX, randY, randZ
t = task.frame
if t > self.spl_u.max():
print 'End of Spline, End task'
# fName = 'trace__' + self.pathGen.__name__ + '.npz'
# np.savez( fName, loss=self.loss_ary, gt=self.gt_ary, pred=self.pred_ary )
# print 'PathData File Written : ', fName
# print 'Visualize result : `python tools/analyse_path_trace_subplot.py', fName, '`'
return None
#
# set pose in each camera
# Note: The texture is grided images in a col-major format
# TODO : since it is going to be only 1 camera eliminate this loop to simply code
poses = np.zeros( (len(self.cameraList), 4), dtype='float32' )
for i in range(len(self.cameraList)): #here usually # of cams will be 1 (for TestRenderer)
indx = TestRenderer.renderIndx
pt = interpolate.splev( t, self.spl_tck)
#randX,randY, randZ, randYaw, randPitch, randRoll = self.monte_carlo_sample()
randX = pt[0]
randY = pt[1]
randZ = pt[2]
randYaw = pt[3]
randPitch = 0
randRoll = 0
self.cameraList[i].setPos(randX,randY,randZ)
self.cameraList[i].setHpr(randYaw,-90+randPitch,0+randRoll)
poses[i,0] = randX
poses[i,1] = randY
poses[i,2] = randZ
poses[i,3] = randYaw
# make note of the poses just set as this will take effect next
if TestRenderer.renderIndx == 0:
TestRenderer.renderIndx = TestRenderer.renderIndx + 1
# self.putBoxes(0,0,0, scale=100)
self.prevPoses = poses
return task.cont
#
# Retrive Rendered Data
tex = self.win2.getScreenshot()
# A = np.array(tex.getRamImageAs("RGB")).reshape(960,1280,3) #@#
A = np.array(tex.getRamImageAs("RGB")).reshape(240,320,3)
# A = np.zeros((960,1280,3))
# A_bgr = cv2.cvtColor(A.astype('uint8'),cv2.COLOR_RGB2BGR)
# cv2.imwrite( str(TestRenderer.renderIndx)+'.png', A_bgr.astype('uint8') )
# myTexture = self.win2.getTexture()
# print myTexture
# retrive poses from prev render
texPoses = self.prevPoses
#
# Cut rendered data into individual image. Note rendered data will be 4X4 grid of images
#960 rows and 1280 cols (4x4 image-grid)
nRows = 240
nCols = 320
# Iterate over the rendered texture in a col-major format
c=0
# TODO : Eliminate this 2-loop as we know there is only 1 display region
#if self.q_imStack.qsize() < 150: #no limit on queue size
# for j in range(1): #j is for cols-indx
# for i in range(1): #i is for rows-indx
i=0
j=0
#print i*nRows, j*nCols, (i+1)*nRows, (j+1)*nCols
im = A[i*nRows:(i+1)*nRows,j*nCols:(j+1)*nCols,:]
#imX = im.astype('float32')/255. - .5 # TODO: have a mean image
#imX = (im.astype('float32') - 128.0) /128.
imX = im.astype('float32') #- self.meanImage
# Put imX into the queue
# do not queue up if queue size begin to exceed 150
self.q_imStack.put( imX )
self.q_labelStack.put( texPoses[c,:] )
self.putBoxes( texPoses[c,0], texPoses[c,1], texPoses[c,2] )
# print 'putBoxes', texPoses[c,0], texPoses[c,1], texPoses[c,2]
# fname = '__'+str(poses[c,0]) + '_' + str(poses[c,1]) + '_' + str(poses[c,2]) + '_' + str(poses[c,3]) + '_'
# cv2.imwrite( str(TestRenderer.renderIndx)+'__'+str(i)+str(j)+fname+'.png', imX.astype('uint8') )
c = c + 1
#
# Prep for Next Iteration
TestRenderer.renderIndx = TestRenderer.renderIndx + 1
self.prevPoses = poses
# if( TestRenderer.renderIndx > 5 ):
# return None
return task.cont
def step(self):
self.taskMgr.step()
# print 'Queues Status (imStack=%d,labelStack=%d)' %(self.q_imStack.qsize(), self.q_labelStack.qsize())
# Dequeue 1 elements
if self.q_imStack.qsize() > 2: # Do not dequeue if the queue size is less than 2
im = copy.deepcopy( self.q_imStack.get() ) #240x320x3 RGB
y = copy.deepcopy( self.q_labelStack.get() )
return im, y
else:
return None, None
def __init__(self, pathGen=None ):
ShowBase.__init__(self)
self.taskMgr.add( self.renderNtestTask, "renderNtestTask" ) #changing camera poses
self.taskMgr.add( self.putAxesTask, "putAxesTask" ) #draw co-ordinate axis
# Misc Setup
self.render.setAntialias(AntialiasAttrib.MAuto)
self.setFrameRateMeter(True)
self.tcolor = TerminalColors.bcolors()
#
# Set up Mesh (including load, position, orient, scale)
self.setupMesh()
self.positionMesh()
# Custom Render
# Important Note: self.render displays the low_res and self.scene0 is the images to retrive
self.scene0 = NodePath("scene0")
# cytX = copy.deepcopy( cyt )
self.low_res.reparentTo(self.render)
self.cyt.reparentTo(self.scene0)
self.cyt2.reparentTo(self.scene0)
#
# Make Buffering Window
bufferProp = FrameBufferProperties().getDefault()
props = WindowProperties()
# props.setSize(1280, 960)
props.setSize(320, 240) #@#
win2 = self.graphicsEngine.makeOutput( pipe=self.pipe, name='wine1',
sort=-1, fb_prop=bufferProp , win_prop=props,
flags=GraphicsPipe.BFRequireWindow)
#flags=GraphicsPipe.BFRefuseWindow)
# self.window = win2#self.win #dr.getWindow()
self.win2 = win2
# self.win2.setupCopyTexture()
# Adopted from : https://www.panda3d.org/forums/viewtopic.php?t=3880
#
# Set Multiple Cameras
self.cameraList = []
# for i in range(4*4):
for i in range(1*1): #@#
print 'Create camera#', i
self.cameraList.append( self.customCamera( str(i) ) )
# Disable default camera
# dr = self.camNode.getDisplayRegion(0)
# dr.setActive(0)
#
# Set Display Regions (4x4)
dr_list = self.customDisplayRegion(1,1)
#
# Setup each camera
for i in range(len(dr_list)):
dr_list[i].setCamera( self.cameraList[i] )
#
# Set buffered Queues (to hold rendered images and their positions)
# each queue element will be an RGB image of size 240x320x3
self.q_imStack = Queue.Queue()
self.q_labelStack = Queue.Queue()
#
# Setting up Splines
# Note: Start interpolation at 50,
if pathGen is None:
# self.pathGen = PathMaker.PathMaker().path_flat_h
# self.pathGen = PathMaker.PathMaker().path_smallM
# self.pathGen = PathMaker.PathMaker().path_yaw_only
# self.pathGen = PathMaker.PathMaker().path_bigM
# self.pathGen = PathMaker.PathMaker().path_flat_spiral
# self.pathGen = PathMaker.PathMaker().path_helix
# self.pathGen = PathMaker.PathMaker().path_like_real
# self.pathGen = PathMaker.PathMaker().path_like_real2
self.pathGen = PathMaker.PathMaker().path_large_loop
else:
self.pathGen = pathGen
t,X = self.pathGen()
self.spl_tck, self.spl_u = interpolate.splprep(X.T, u=t.T, s=0.0, per=1)
print 'Test Renderer Init Done'
print self.tcolor.OKGREEN, 'Test Renderer Init Done', self.tcolor.ENDC
# Setup NetVLAD Renderer - This renderer is custom made for NetVLAD training
# It renders 16 images at a time. (q, (P1,P2,..P5), (N1,N2,...,N10)).
# ie. 1st image is im, next 5 are near this im (potential positives).
# Last 10 are far from im (definite negatives)
class NetVLADRenderer(ShowBase):
renderIndx=0
# Basic Mesh & Camera Setup
def loadAllTextures(self, mesh, basePath, silent=True):
""" Loads texture files for a mesh """
c = 0
for child in mesh.getChildren():
submesh_name = child.get_name()
submesh_texture = basePath + submesh_name[:-5] + 'tex0.jpg'
child.setTexture( self.loader.loadTexture(submesh_texture) )
if silent == False:
print 'Loading texture file : ', submesh_texture
c = c + 1
print self.tcolor.OKGREEN, "Loaded ", c, "textures", self.tcolor.ENDC
def setupMesh(self):
""" Loads the .obj files. Will load mesh sub-divisions separately """
print 'Attempt Loading Mesh VErtices, FAces'
self.cyt = self.loader.loadModel( 'model_l/l6/level_6_0_0.obj' )
self.cyt2 = self.loader.loadModel( 'model_l/l6/level_6_128_0.obj' )
self.low_res = self.loader.loadModel( 'model_l/l3/level_3_0_0.obj' )
print self.tcolor.OKGREEN, 'Done Loading Vertices', self.tcolor.ENDC
print 'Attempt Loading Textures'
self.loadAllTextures( self.cyt, 'model_l/l6/')
self.loadAllTextures( self.cyt2, 'model_l/l6/')
self.loadAllTextures( self.low_res, 'model_l/l3/')
print self.tcolor.OKGREEN, 'Done Loading Textures', self.tcolor.ENDC
def positionMesh(self):
""" WIll have to manually adjust this for ur mesh. I position the
center where I fly my drone and oriented in ENU (East-north-up)
cords for easy alignment of GPS and my cordinates. If your model
is not metric scale will have to adjust for that too"""
self.cyt.setPos( 140,-450, 150 )
self.cyt2.setPos( 140,-450, 150 )
self.low_res.setPos( 140,-450, 150 )
self.cyt.setHpr( 198, -90, 0 )
self.cyt2.setHpr( 198, -90, 0 )
self.low_res.setHpr( 198, -90, 0 )
self.cyt.setScale(172)
self.cyt2.setScale(172)
self.low_res.setScale(172)
def customCamera(self, nameIndx):
lens = self.camLens
lens.setFov(83)
print 'self.customCamera : Set FOV at 83'
my_cam = Camera("cam"+nameIndx, lens)
my_camera = self.scene0.attachNewNode(my_cam)
# my_camera = self.render.attachNewNode(my_cam)
my_camera.setName("camera"+nameIndx)
return my_camera
def customDisplayRegion(self, rows, cols):
rSize = 1.0 / rows
cSize = 1.0 / cols
dr_list = []
for i in range(0,rows):
for j in range(0,cols):
# print i*rSize, (i+1)*rSize, j*cSize, (j+1)*cSize
dr_i = self.win2.makeDisplayRegion(i*rSize, (i+1)*rSize, j*cSize, (j+1)*cSize)
dr_i.setSort(-5)
dr_list.append( dr_i )
return dr_list
def mc_default( self, cam0X, cam0Y, cam0Z ):
return 0,0,80,0,0,0
def mc_far( self, cam0X, cam0Y, cam0Z ):
rf = np.random.uniform
nZ = rf(self.mc_Z_min,self.mc_Z_max)
fov = 1.3962 #80 degrees
sigma = cam0Z * np.tan(fov/2.)/3
nX = rf(self.mc_X_min, cam0X - 2*sigma) if rf(-1,1) > 0 else rf(cam0X + 2*sigma, self.mc_X_max )
nY = rf(self.mc_Y_min, cam0Y - 2*sigma) if rf(-1,1) > 0 else rf(cam0Y + 2*sigma, self.mc_Y_max )
yaw = rf(self.mc_yaw_min, self.mc_yaw_max)
return nX, nY, nZ, yaw, 0 , 0
# Return a random sample near (cam0X,cam0Y,cam0Z)
def mc_near( self, cam0X, cam0Y, cam0Z ):
rf = np.random.uniform
nZ = rf(self.mc_Z_min,self.mc_Z_max)
fov = 1.3962 #80 degrees
sigma = cam0Z * np.tan(fov/2.)/3
yaw = rf(self.mc_yaw_min, self.mc_yaw_max)
return rf(cam0X-sigma,cam0X+sigma), rf(cam0Y-sigma,cam0Y+sigma),nZ,yaw,0.,0.
## Gives a random 6-dof pose. Need to set params manually here.
## X,Y,Z, Yaw(abt Z-axis), Pitch(abt X-axis), Roll(abt Y-axis)
## @param No
def monte_carlo_sample(self):
# mc_X_min etc are set in constructor
X = np.random.uniform(self.mc_X_min,self.mc_X_max)
Y = np.random.uniform(self.mc_Y_min,self.mc_Y_max)
Z = np.random.uniform(self.mc_Z_min,self.mc_Z_max)
yaw = np.random.uniform( self.mc_yaw_min, self.mc_yaw_max)
roll = 0#np.random.uniform( self.mc_roll_min, self.mc_roll_max)
pitch = 0#np.random.uniform( self.mc_pitch_min, self.mc_pitch_max)