-
Notifications
You must be signed in to change notification settings - Fork 84
/
createosmanomaly.py
500 lines (361 loc) · 18.4 KB
/
createosmanomaly.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
import sys
sys.path.append("Mask_RCNN")
import os
import sys
import glob
import osmmodelconfig
import skimage
import math
import imagestoosm.config as osmcfg
import model as modellib
import visualize as vis
import numpy as np
import csv
import QuadKey.quadkey as quadkey
import shapely.geometry as geometry
import shapely.affinity as affinity
import matplotlib.pyplot as plt
import cv2
import scipy.optimize
import time
from skimage import draw
from skimage import io
showFigures = False
def toDegrees(rad):
return rad * 180/math.pi
def writeOSM( osmFileName,featureName, simpleContour,tilePixel, qkRoot) :
with open(osmFileName,"wt",encoding="ascii") as f:
f.write("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n")
f.write("<osm version=\"0.6\">\n")
id = -1
for pt in simpleContour :
geo = quadkey.TileSystem.pixel_to_geo( (pt[0,0]+tilePixel[0],pt[0,1]+tilePixel[1]),qkRoot.level)
f.write(" <node id=\"{}\" lat=\"{}\" lon=\"{}\" />\n".format(id,geo[0],geo[1]))
id -= 1
f.write(" <way id=\"{}\" visible=\"true\">\n".format(id))
id = -1
for pt in simpleContour :
f.write(" <nd ref=\"{}\" />\n".format(id))
id -= 1
f.write(" <nd ref=\"{}\" />\n".format(-1))
f.write(" <tag k=\"{}\" v=\"{}\" />\n".format("leisure","pitch"))
f.write(" <tag k=\"{}\" v=\"{}\" />\n".format("sport",featureName))
f.write(" </way>\n")
f.write("</osm>\n")
f.close
def writeShape(wayNumber, finalShape, image, bbTop,bbHeight,bbLeft,bbWidth) :
nPts = int(finalShape.length)
if ( nPts > 5000) :
nPts = 5000
fitContour = np.zeros((nPts,1,2), dtype=np.int32)
if ( nPts > 3):
for t in range(0,nPts) :
pt = finalShape.interpolate(t)
fitContour[t,0,0] = pt.x
fitContour[t,0,1] = pt.y
fitContour = [ fitContour ]
fitContour = [ cv2.approxPolyDP(cnt,2,True) for cnt in fitContour]
image = np.copy(imageNoMasks)
cv2.drawContours(image, fitContour,-1, (0,255,0), 2)
if ( showFigures ):
fig.add_subplot(2,2,3)
plt.title(featureName + " " + str(r['scores'][i]) + " Fit")
plt.imshow(image[bbTop:bbTop+bbHeight,bbLeft:bbLeft+bbWidth])
while ( os.path.exists( "anomaly/add/{0:06d}.osm".format(wayNumber) )) :
wayNumber += 1
debugFileName = os.path.join( inference_config.ROOT_DIR, "anomaly","add","{0:06d}.jpg".format(wayNumber))
io.imsave(debugFileName,image[bbTop:bbTop+bbHeight,bbLeft:bbLeft+bbWidth],quality=100)
osmFileName = os.path.join( inference_config.ROOT_DIR, "anomaly","add","{0:06d}.osm".format(wayNumber))
writeOSM( osmFileName,featureName, fitContour[0],tilePixel, qkRoot)
if (showFigures ):
plt.show(block=False)
plt.pause(0.05)
return wayNumber
ROOT_DIR_ = os.path.dirname(os.path.realpath(sys.argv[0]))
MODEL_DIR = os.path.join(ROOT_DIR_, "logs")
class InferenceConfig(osmmodelconfig.OsmModelConfig):
GPU_COUNT = 1
IMAGES_PER_GPU = 1
ROOT_DIR = ROOT_DIR_
inference_config = InferenceConfig()
fullTrainingDir = os.path.join( ROOT_DIR_, osmcfg.trainDir,"*")
fullImageList = []
for imageDir in glob.glob(fullTrainingDir):
if ( os.path.isdir( os.path.join( fullTrainingDir, imageDir) )):
id = os.path.split(imageDir)[1]
fullImageList.append( id)
# Training dataset
dataset_full = osmmodelconfig.OsmImagesDataset(ROOT_DIR_)
dataset_full.load(fullImageList, inference_config.IMAGE_SHAPE[0], inference_config.IMAGE_SHAPE[1])
dataset_full.prepare()
inference_config.display()
# Recreate the model in inference mode
model = modellib.MaskRCNN(mode="inference",
config=inference_config,
model_dir=MODEL_DIR)
# Get path to saved weights
# Either set a specific path or find last trained weights
# model_path = os.path.join(ROOT_DIR, ".h5 file name here")
model_path = model.find_last()[1]
print(model_path)
# Load trained weights (fill in path to trained weights here)
assert model_path != "", "Provide path to trained weights"
print("Loading weights from ", model_path)
model.load_weights(model_path, by_name=True)
print("Reading in OSM data")
# load up the OSM features into hash of arrays of polygons, in pixels
features = {}
for classDir in os.listdir(osmcfg.rootOsmDir) :
classDirFull = os.path.join( osmcfg.rootOsmDir,classDir)
for fileName in os.listdir(classDirFull) :
fullPath = os.path.join( osmcfg.rootOsmDir,classDir,fileName)
with open(fullPath, "rt") as csvfile:
csveader = csv.reader(csvfile, delimiter='\t')
pts = []
for row in csveader:
latLot = (float(row[0]),float(row[1]))
pixel = quadkey.TileSystem.geo_to_pixel(latLot,osmcfg.tileZoom)
pts.append(pixel)
feature = {
"geometry" : geometry.Polygon(pts),
"filename" : fullPath
}
if ( (classDir in features) == False) :
features[classDir] = []
features[classDir].append( feature )
# make the output dirs, a fresh start is possible just by deleting anomaly
if ( not os.path.isdir("anomaly")) :
os.mkdir("anomaly")
if ( not os.path.isdir("anomaly/add")) :
os.mkdir("anomaly/add")
if ( not os.path.isdir("anomaly/replace")) :
os.mkdir("anomaly/replace")
if ( not os.path.isdir("anomaly/overlap")) :
os.mkdir("anomaly/overlap")
fig = {}
if ( showFigures):
fig = plt.figure()
wayNumber = 0
startTime = time.time()
count = 1
for image_index in dataset_full.image_ids :
currentTime = time.time()
howLong = currentTime-startTime
secPerImage = howLong/count
imagesLeft = len(dataset_full.image_ids)-count
timeLeftHrs = (imagesLeft*secPerImage)/3600.0
print("Processing {} of {} {:2.1f} hrs left".format(count,len(dataset_full.image_ids),timeLeftHrs))
count += 1
image, image_meta, gt_class_id, gt_bbox, gt_mask = modellib.load_image_gt(dataset_full, inference_config,image_index, use_mini_mask=False)
info = dataset_full.image_info[image_index]
# get the pixel location for this training image.
metaFileName = os.path.join( inference_config.ROOT_DIR, osmcfg.trainDir,info['id'],info['id']+".txt")
quadKeyStr = ""
with open(metaFileName) as metafile:
quadKeyStr = metafile.readline()
quadKeyStr = quadKeyStr.strip()
qkRoot = quadkey.from_str(quadKeyStr)
tilePixel = quadkey.TileSystem.geo_to_pixel(qkRoot.to_geo(), qkRoot.level)
# run the network
results = model.detect([image], verbose=0)
r = results[0]
maxImageSize = 256*3
featureMask = np.zeros((maxImageSize, maxImageSize), dtype=np.uint8)
pts = []
pts.append( ( tilePixel[0]+0,tilePixel[1]+0 ) )
pts.append( ( tilePixel[0]+0,tilePixel[1]+maxImageSize ) )
pts.append( ( tilePixel[0]+maxImageSize,tilePixel[1]+maxImageSize ) )
pts.append( ( tilePixel[0]+maxImageSize,tilePixel[1]+0 ) )
imageBoundingBoxPoly = geometry.Polygon(pts)
foundFeatures = {}
for featureType in osmmodelconfig.featureNames.keys() :
foundFeatures[featureType ] = []
for feature in features[featureType] :
if ( imageBoundingBoxPoly.intersects( feature['geometry']) ) :
xs, ys = feature['geometry'].exterior.coords.xy
outOfRangeCount = len([ x for x in xs if x < tilePixel[0] or x >= tilePixel[0]+maxImageSize ])
outOfRangeCount += len([ y for y in ys if y < tilePixel[1] or y >= tilePixel[1]+maxImageSize ])
if ( outOfRangeCount == 0) :
foundFeatures[featureType ].append( feature)
# draw black lines showing where osm data is
for featureType in osmmodelconfig.featureNames.keys() :
for feature in foundFeatures[featureType] :
xs, ys = feature['geometry'].exterior.coords.xy
xs = [ x-tilePixel[0] for x in xs]
ys = [ y-tilePixel[1] for y in ys]
rr, cc = draw.polygon_perimeter(xs,ys,(maxImageSize,maxImageSize))
image[cc,rr] = 0
imageNoMasks = np.copy(image)
for i in range( len(r['class_ids'])) :
mask = r['masks'][:,:,i]
edgePixels = 15
outside = np.sum( mask[0:edgePixels,:]) + np.sum( mask[-edgePixels:-1,:]) + np.sum( mask[:,0:edgePixels]) + np.sum( mask[:,-edgePixels:-1])
image = np.copy(imageNoMasks)
if ( r['scores'][i] > 0.98 and outside == 0 ) :
featureFound = False
for featureType in osmmodelconfig.featureNames.keys() :
for feature in foundFeatures[featureType] :
classId = osmmodelconfig.featureNames[featureType]
if ( classId == r['class_ids'][i] ) :
xs, ys = feature['geometry'].exterior.coords.xy
xs = [ x-tilePixel[0] for x in xs]
ys = [ y-tilePixel[1] for y in ys]
xsClipped = [ min( max( x,0),maxImageSize) for x in xs]
ysClipped = [ min( max( y,0),maxImageSize) for y in ys]
featureMask.fill(0)
rr, cc = draw.polygon(xs,ys,(maxImageSize,maxImageSize))
featureMask[cc,rr] = 1
maskAnd = featureMask * mask
overlap = np.sum(maskAnd )
if ( outside == 0 and overlap > 0) :
featureFound = True
if ( featureFound == False) :
weight = 0.25
# get feature name
featureName = ""
for featureType in osmmodelconfig.featureNames.keys() :
if ( osmmodelconfig.featureNames[featureType] == r['class_ids'][i] ) :
featureName = featureType
#if ( r['class_ids'][i] == 1):
# vis.apply_mask(image,mask,[weight,0,0])
#if ( r['class_ids'][i] == 2):
# vis.apply_mask(image,mask,[weight,weight,0])
#if ( r['class_ids'][i] == 3):
# vis.apply_mask(image,mask,[0.0,0,weight])
mask = mask.astype(np.uint8)
mask = mask * 255
ret,thresh = cv2.threshold(mask,127,255,0)
im2, rawContours,h = cv2.findContours(thresh,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_NONE)
bbLeft,bbTop,bbWidth,bbHeight = cv2.boundingRect(rawContours[0])
bbBuffer = 75
bbLeft = max(bbLeft-bbBuffer,0)
bbRight = min(bbLeft+2*bbBuffer+bbWidth,maxImageSize)
bbWidth = bbRight-bbLeft
bbTop = max(bbTop-bbBuffer,0)
bbBottom = min(bbTop+2*bbBuffer+bbHeight,maxImageSize-1)
bbHeight = bbBottom-bbTop
image = np.copy(imageNoMasks)
cv2.drawContours(image, rawContours,-1, (0,255,0), 2)
if ( showFigures ):
fig.add_subplot(2,2,1)
plt.title(featureName + " " + str(r['scores'][i]) + " Raw")
plt.imshow(image[bbTop:bbTop+bbHeight,bbLeft:bbLeft+bbWidth])
simpleContour = [ cv2.approxPolyDP(cnt,5,True) for cnt in rawContours]
image = np.copy(imageNoMasks)
cv2.drawContours(image, simpleContour,-1, (0,255,0), 2)
if ( showFigures ):
fig.add_subplot(2,2,2)
plt.title(featureName + " " + str(r['scores'][i]) + " Simplify")
plt.imshow(image[bbTop:bbTop+bbHeight,bbLeft:bbLeft+bbWidth])
simpleContour = simpleContour[0]
print(" {}".format(featureName))
if ( featureName == "baseball" and isinstance(simpleContour,np.ndarray) ):
while ( os.path.exists( "anomaly/add/{0:06d}.osm".format(wayNumber) )) :
wayNumber += 1
debugFileName = os.path.join( inference_config.ROOT_DIR, "anomaly","add","{0:06d}.jpg".format(wayNumber))
io.imsave(debugFileName,image[bbTop:bbTop+bbHeight,bbLeft:bbLeft+bbWidth],quality=100)
osmFileName = os.path.join( inference_config.ROOT_DIR, "anomaly","add","{0:06d}.osm".format(wayNumber))
writeOSM( osmFileName,featureName, simpleContour,tilePixel, qkRoot)
fitContour = simpleContour
if ( featureName == 'baseball' ) :
def makePie(paramsX):
centerX,centerY,width,angle = paramsX
pts = []
pts.append((0,0))
pts.append((width,0))
step = math.pi/10
r = step
while r < math.pi/2:
x = math.cos(r)*width
y = math.sin(r)*width
pts.append( (x,y) )
r += step
pts.append( (0,width))
pts.append( (0,0))
fitShape = geometry.LineString(pts)
fitShape = affinity.translate(fitShape, -width/2,-width/2 )
fitShape = affinity.rotate(fitShape,angle )
fitShape = affinity.translate(fitShape, centerX,centerY )
return fitShape
def fitPie(paramsX):
fitShape = makePie(paramsX)
huberCutoff = 5
sum = 0
for cnt in rawContours:
for pt in cnt:
p = geometry.Point(pt[0])
d = p.distance(fitShape)
if ( d < huberCutoff) :
sum += 0.5 * d * d
else:
sum += huberCutoff*(math.fabs(d)-0.5*huberCutoff)
return sum
cm = np.mean( rawContours[0],axis=0)
results = []
angleStepCount = 8
for angleI in range(angleStepCount):
centerX = cm[0,0]
centerY = cm[0,1]
width = math.sqrt(cv2.contourArea(rawContours[0]))
angle = 360 * float(angleI)/angleStepCount
x0 = np.array([centerX,centerY,width,angle ])
resultR = scipy.optimize.minimize(fitPie, x0, method='nelder-mead', options={'xtol': 1e-6,'maxiter':50 })
results.append(resultR)
bestScore = 1e100
bestResult = {}
for result in results:
if result.fun < bestScore :
bestScore = result.fun
bestResult = result
bestResult = scipy.optimize.minimize(fitPie, bestResult.x, method='nelder-mead', options={'xtol': 1e-6 })
finalShape = makePie(bestResult.x)
wayNumber = writeShape(wayNumber, finalShape, image, bbTop,bbHeight,bbLeft,bbWidth)
for result in results:
angle = result.x[3]
angleDelta = int(math.fabs(result.x[3]-bestResult.x[3])) % 360
if result.fun < 1.2*bestScore and angleDelta > 45 :
result = scipy.optimize.minimize(fitPie, result.x, method='nelder-mead', options={'xtol': 1e-6 })
finalShape = makePie(result.x)
wayNumber = writeShape(wayNumber, finalShape, image, bbTop,bbHeight,bbLeft,bbWidth)
else:
def makeRect(paramsX):
centerX,centerY,width,height,angle = paramsX
pts = [
(-width/2,height/2),
(width/2,height/2),
(width/2,-height/2),
(-width/2,-height/2),
(-width/2,height/2)]
fitShape = geometry.LineString(pts)
fitShape = affinity.rotate(fitShape, angle,use_radians=True )
fitShape = affinity.translate(fitShape, centerX,centerY )
return fitShape
def fitRect(paramsX):
fitShape = makeRect(paramsX)
sum = 0
for cnt in rawContours:
for pt in cnt:
p = geometry.Point(pt[0])
d = p.distance(fitShape)
sum += d*d
return sum
cm = np.mean( rawContours[0],axis=0)
result = {}
angleStepCount = 8
for angleI in range(angleStepCount):
centerX = cm[0,0]
centerY = cm[0,1]
width = math.sqrt(cv2.contourArea(rawContours[0]))
height = width
angle = 2*math.pi * float(angleI)/angleStepCount
x0 = np.array([centerX,centerY,width,height,angle ])
resultR = scipy.optimize.minimize(fitRect, x0, method='nelder-mead', options={'xtol': 1e-6,'maxiter':50 })
if ( angleI == 0):
result = resultR
if ( resultR.fun < result.fun):
result = resultR
#print("{} {}".format(angle * 180.0 / math.pi,resultR.fun ))
resultR = scipy.optimize.minimize(fitRect, resultR.x, method='nelder-mead', options={'xtol': 1e-6 })
#print(result)
finalShape = makeRect(result.x)
wayNumber = writeShape(wayNumber, finalShape, image, bbTop,bbHeight,bbLeft,bbWidth)