-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathretrieval.py
596 lines (456 loc) · 25.6 KB
/
retrieval.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
#!/usr/bin/python
import os,sys,math,operator,cv2,glob,pickle
import numpy as np
from math import radians, cos, sin, asin, sqrt
import similarity
from similarity import Point
from random import shuffle
import scipy.spatial.distance
import IPython
import ImageDraw
import ImageColor
import argparse
try:
from PIL import Image
except:
import Image
import matplotlib.pyplot as plt
parser = argparse.ArgumentParser()
parser.add_argument('--f', type=str, default="frame/weather/train/", help='retrieval directory')
parser.add_argument('--q', type=str, default="cross", help='query')
parser.add_argument('--r', type=int, default=0, help='retrieved accuracy')
parser.add_argument('--w1', type=float, default=0.315, help='weight')
parser.add_argument('--w2', type=float, default=0.437, help='weight')
parser.add_argument('--w3', type=float, default=0.4, help='weight')
args = parser.parse_args()
def demo(img_name, index, im, bb_array, gt, best_index, output_dir):
#painting
fig, ax = plt.subplots(figsize=(12, 12))
ax.imshow(im, aspect='equal')
top = 5
for single_bb in xrange(top):
if single_bb==0:
ax.add_patch(
plt.Rectangle((bb_array[single_bb][0:2])
,bb_array[single_bb][2]-bb_array[single_bb][0]
,bb_array[single_bb][3]-bb_array[single_bb][1]
,fill=False,edgecolor='yellow', linewidth=3.5))
ax.text(bb_array[single_bb][0] + 6, bb_array[single_bb][1] - 14,
'{}, {:.3f}'.format(single_bb+1, float(bb_array[single_bb][4])),
bbox=dict(facecolor='yellow', alpha=0.5),
fontsize=14, color='black')
else:
ax.add_patch(
plt.Rectangle((bb_array[single_bb][0:2])
,bb_array[single_bb][2]-bb_array[single_bb][0]
,bb_array[single_bb][3]-bb_array[single_bb][1]
,fill=False,edgecolor='blue', linewidth=3.5))
ax.text(bb_array[single_bb][0] + 6, bb_array[single_bb][1] - 14,
'{}, {:.3f}'.format(single_bb+1, float(bb_array[single_bb][4])),
bbox=dict(facecolor='blue', alpha=0.5),
fontsize=14, color='white')
ax.add_patch(
plt.Rectangle((gt[0:2])
,gt[2]-gt[0]
,gt[3]-gt[1]
,fill=False,edgecolor='red', linewidth=3.5))
"""
ax.add_patch(
plt.Rectangle((bb_array[best_index][0:2])
,bb_array[best_index][2]-bb_array[best_index][0]
,bb_array[best_index][3]-bb_array[best_index][1]
,fill=False,edgecolor='green', linewidth=3.5))
ax.text(bb_array[best_index][0] + 6, bb_array[best_index][1] - 14,
'{:.3f}'.format(float(bb_array[best_index][4])),
bbox=dict(facecolor='green', alpha=0.5),
fontsize=14, color='white')
"""
ax.set_title(('{}. {}').format(index, img_name),fontsize=14)
plt.axis('off')
plt.tight_layout()
plt.draw()
fig.savefig(os.path.join(output_dir, '{}.jpg'.format(index)),bbox_inches='tight')
"""
top=5
dr = ImageDraw.Draw(im)
dr.rectangle(gt, outline = "red")
for single_bb in xrange(top):
dr.rectangle(bb_array[single_bb][0:4],
outline = (0, 0, 255*(top-single_bb)/5))
#im.thumbnail( (400,100) )
im.save('output/{}.jpg'.format(index))
"""
def main():
# setting
iou_threshold = 0.3
poi_dir = "poi/"
#frame_dir = "old_frame/"
#frame_dir = "frame/weather/test/"
#frame_dir = "frame/location_3/test/"
#frame_dir = "frame/location/test/"
#frame_dir = "frame/all/"
frame_dir = args.f
#bb_dir = "500_bb_gt/"
bb_dir = "faster_bb/"
#bb_dir = "500_bb/"
#search_dir = "search/"
#streetview_dir = "streetview/"
#streetview_dir = "streetview_clean/"
#aerial_dir = "aerial_clean"
#cross_dir = "cross"
#query_list = [search_dir, streetview_dir]
#query_list = [search_dir]
#query_list = [search_dir]
#query_list = ["image_gt"]
query_list = [args.q]
#layer_list = ["pool5", "fc6","fc7"]
layer_list = ["fc6"]
visual_dir = "visual_feature"
visual_cnn = "CNN"
visual_place = "placeCNN"
visual_faster = "fasterRCNN"
visual_triplet = "triplet"
#visual_feature_list = [visual_cnn, visual_place, visual_triplet]
visual_feature_list = [visual_triplet]
output_dir = "output"
if not os.path.exists(output_dir):
os.makedirs(output_dir)
streetview_num = 1
never = 0
frame_info = {}
ans_hit = 0
img_num = 0
img_sum = []
#load frame info: drone directeion
with open("location.txt", 'r') as f:
for line in f:
token = line.strip().split("\t")
frame_info[token[0]] = token[1:5]
frame_info[token[0]] = [float(i) for i in frame_info[token[0]]]
# for random many times
iteration_max = 1
index = 0
if args.r == -1:
recall = range(0, 11)
else:
recall = [args.r]
for rr in recall:
hit = 0
poi_num = 0
for iteration in xrange(iteration_max):
for img_name in os.listdir(frame_dir):
temp = img_name
im = Image.open(os.path.join(frame_dir, img_name))
width, height = im.size
img_name = img_name.replace(".jpg","").replace(".png","")
# preload bb deep feature pkl
bb_proposal = []
bb_proposal_pkl = {}
for layer in layer_list:
for visual_feature in visual_feature_list:
bb_proposal_pkl[visual_feature+"_"+layer] = []
bb_score=[]
proposal_max = 200
proposal_num = 0
with open(bb_dir+img_name+".txt", 'r') as ff:
for linee in ff:
if proposal_num == proposal_max:
break
token = linee.strip().split()
bb = [int(float(token[0])), int(float(token[1])), int(float(token[2])), int(float(token[3]))]
[bb_width,bb_height] = [bb[3]-bb[1],bb[2]-bb[0]]
"""
if bb_height>= height/2 or bb_width>= width/2 or bb_height<= height/50 or bb_width<= width/50:
continue
"""
bb_proposal.append(bb)
bb_score.append(float(token[4]))
file_name = img_name+'_'+str(bb[0])+"_"+str(bb[1])+"_"+str(bb[2])+"_"+str(bb[3])+".pkl"
for layer in layer_list:
for visual_feature in visual_feature_list:
pkl_name = os.path.join(visual_dir, visual_feature, bb_dir, layer, file_name)
try:
pkl = pickle.load(open(pkl_name,'rb'))
except:
IPython.embed()
bb_proposal_pkl[visual_feature+"_"+layer].append(pkl.reshape(np.product(pkl.shape)))
proposal_num += 1
#first
with open(poi_dir+img_name+".txt", 'r') as f:
center_pos = [0,0]
[center_pos[0],center_pos[1],yaw,frame_height] = frame_info[img_name]
easy_num = 0
matching_pair=[]
ans_gt=[]
ans_list=[]
ans_index=0
img_num += 1
gt_list = []
for line in f:
token = line.strip().split("\t")
index+= 1
#print(index)
#print(img_name)
"""
easy_num+= 1
if easy_num>7:
break
"""
im = Image.open(frame_dir+temp)
gps_pos = [0, 0]
[name,gps_pos[0],gps_pos[1],google_type,img_ref,gt] = token
str_gps_pos = [gps_pos[0],gps_pos[1]]
query_name = str(str_gps_pos[0])+'_'+str(str_gps_pos[1])
gps_pos[0] = float(gps_pos[0])
gps_pos[1] = float(gps_pos[1])
gt = gt.split(',')
gt = [float(i) for i in gt]
ans_gt.append(gt)
gt_list.append([gt, gps_pos, center_pos, yaw])
max_dis = sqrt((height/2)**2+width**2)
bb_array = []
best_bb = []
score = []
#map_middle = [math.tan(1*math.pi/180*similarity.yawtoangle(yaw)),1]
#map_vector = [gps_pos[0]-center_pos[0],gps_pos[1]-center_pos[1]]
#map_angle = math.degrees(similarity.angle(map_middle,map_vector))
map_middle = similarity.yawtoangle(yaw)
map_vector = math.degrees(similarity.angle([0,1],
similarity.gpsvecotr(Point(gps_pos[1], gps_pos[0]),
Point(center_pos[1], center_pos[0]))))
#map_vector = math.degrees(similarity.angle([0,1],[gps_pos[1]-center_pos[1],gps_pos[0]-center_pos[0]]))
#IPython.embed()
if map_vector<0:
map_vector = -map_vector
else:
map_vector = 360-map_vector
map_angle = map_middle-map_vector
drone_middle = [0,1]
visual_dis_list = []
for query in query_list:
for layer in layer_list:
for visual_feature in visual_feature_list:
target_pkl = []
# origin
pkl_name = os.path.join(visual_dir, visual_feature, query, layer, query_name+".pkl")
# for gt testing
#pkl_name = os.path.join(visual_dir, visual_feature, query, layer, img_name+"_"+query_name+".pkl")
pkl = pickle.load(open(pkl_name,'rb'))
bb_proposal_array = np.array(bb_proposal_pkl[visual_feature+"_"+layer])
target_pkl.append(pkl.reshape(np.product(pkl.shape)))
target_pkl = np.array(target_pkl)
visual_dis_list.append(scipy.spatial.distance.cdist(target_pkl,bb_proposal_array))
for bb_index in xrange(len(bb_proposal)):
bb = bb_proposal[bb_index]
#im_crop = im.crop((bb[0],bb[1],bb[2],bb[3]))
#data_histo = get_hsv_histo(im_crop)
#dis = color_dis(query_histo,data_histo)
drone_vector = [(bb[0]+bb[2])/2-width/2,height-bb[3]]
drone_angle = math.degrees(similarity.angle(drone_middle, drone_vector))
#distance
angle_dis = math.fabs(map_angle-drone_angle)
"""
if angle_dis>45:
angle_dis=400
"""
scale_dis = similarity.likelyscale(bb,height,width)
iou = similarity.iou(bb, gt)
p1 = Point(gps_pos[1], gps_pos[0])
origin_p = Point(center_pos[1], center_pos[0])
if iou >= iou_threshold:
distance_x_dis = similarity.likelydistance(gps_pos, center_pos, bb, height, width,
similarity.disofpoint2line(p1, origin_p, similarity.yawtovector(yaw)), True,True)
distance_y_dis = similarity.likelydistance(gps_pos, center_pos, bb, height, width,
similarity.disofpoint2line(p1, origin_p, similarity.yawtovector(yaw)), False,True)
else:
distance_x_dis = similarity.likelydistance(gps_pos,center_pos,bb, height, width,
similarity.disofpoint2line(p1, origin_p, similarity.yawtovector(yaw)), True)
distance_y_dis = similarity.likelydistance(gps_pos,center_pos,bb, height, width,
similarity.disofpoint2line(p1, origin_p, similarity.yawtovector(yaw)), False)
visual_dis = []
for visual_single in visual_dis_list:
visual_dis.append(visual_single[0, bb_index])
#visual_dis = similarity.likelyvisual_multiple(target_pkl,bb_proposal_pkl[bb_index])
score.append([angle_dis,
#scale_dis,
distance_x_dis,
distance_y_dis
]
+[visual_single for visual_single in visual_dis]
)
bb_array.append([bb[0],bb[1],bb[2],bb[3]])
#bb_array.append([bb[0],bb[1],bb[2],bb[3],scale_dis+angle_dis+distance_dis])
#print(str(bb)+" "+str(angle_dis)+" "+str(scale_dis)+" "+str(distance_dis)+" "+str(visual_dis))
#print(str(bb)+" "+str(visual_dis))
#bb_array.append([bb[0],bb[1],bb[2],bb[3],dis+likelyscale(bb,height,width)/5+angle_dis/1000+likelydistance(gps_pos,center_pos,bb,height,width,max_dis)])
#print(str(bb)+" "+str(dis)+" "+str(likelyscale(bb,height,width)/5)+" "+str(angle_dis/1000)+" "+str(likelydistance(gps_pos,center_pos,bb,height,width,max_dis)))
#similarity
sim_num = len(score[0])
sim = []
score = np.array(score)
for sim_i in xrange(sim_num):
sim.append(similarity.dis_to_sim(score[:,sim_i]))
for single_bb in xrange(len(bb_array)):
sim_score = 0
for sim_i in xrange(sim_num):
#weight
#all 0.31, 0.44, 0.74
#single 0.31, 0.28, 0.67
#dis 0.28, 0.6
if sim_i == 0:
sim_score += args.w1*sim[sim_i][single_bb] #args.w #0.31
elif sim_i == 1:
sim_score += args.w2*sim[sim_i][single_bb] #0.28
elif sim_i == 2:
sim_score += args.w3*sim[sim_i][single_bb] #0.67
else:
sim_score += sim[sim_i][single_bb]
# filter angle
"""
if sim[sim_i][single_bb]==0:
sim_score-=10000
else:
sim_score+= sim[sim_i][single_bb]
"""
#sim_score += sim[sim_i][single_bb]
#print sim[sim_i][single_bb],
#print
#sim_score += bb_score[single_bb] # add classification score
bb_array[single_bb].append(sim_score)
#bb_array.sort(key = similarity.compare,reverse = True) #cos_sim
for single_bb in xrange(len(bb_array)):
matching_pair.append([ans_index,single_bb,bb_array[single_bb][0:4],bb_array[single_bb][4]])
ans_list.append(ans_index)
ans_index+=1
bb_array.sort(key = lambda x: float(x[4]),reverse = True) #l2_distance
#bb_array.sort(key = lambda x: float(x[4])) #l2_distance
"""
print(bb_array[0][0:4])
print("drone_angle: "+str(bb_array[0][5]))
print("map_middle: "+str(map_middle))
print("map_vector: "+str(map_vector))
print("map_angle: "+str(map_angle))
print(gps_pos)
print(center_pos)
print(yaw)
im.save('test.png')
exit(0)
"""
#shuffle(bb_array)
#print(str(bb_array[0])+" "+str(bb_array[0][4]))
#print(bb_array[0][0:4],gt)
#eval_angle
#hit+= similarity.eval_angle(width,height,bb_array[0][0:4],gt)
best_iou = 0
for bb_index in xrange(len(bb_array)):
iou = similarity.iou(bb_array[bb_index][0:4],gt)
if iou>best_iou:
best_iou = iou
best_index = bb_index
if best_iou < iou_threshold:
never += 1
best_iou = 0.01
#continue
#iou
poi_num+= 1
for bb_index in xrange(len(bb_array)):
iou = similarity.iou(bb_array[bb_index][0:4],gt)
if iou >= iou_threshold and rr == 0:
hit+= float(1) / (bb_index+1)
break
elif iou >= iou_threshold and bb_index < rr:
hit+= float(1)
break
#demo
#demo(img_name, index, im, bb_array, gt, best_index, output_dir)
#gt_list.append(gt, gps_pos, center_pos, yaw)
"""
gt_score = []
for single in gt_list:
[gt, gps_pos, center_pos, yaw] = single
p1 = Point(gps_pos[1], gps_pos[0])
origin_p = Point(center_pos[1], center_pos[0])
gt_score.append([float(gt[2]+gt[0])/2
, similarity.disofpoint2line(p1, origin_p, similarity.yawtovector(yaw))])
gt_score.sort(key = lambda x: float(x[1]), reverse = True)
dcg_list = [gs[0] for gs in gt_score]
print(similarity.DCG(dcg_list))
"""
#IPython.embed()
"""
has_choose=[]
matching_pair.sort(key = lambda x: float(x[3]),reverse = True)
ans_length=[1 for i in xrange(len(ans_gt))]
for pair_index in xrange(len(matching_pair)):
if matching_pair[pair_index][0] in ans_list and not matching_pair[pair_index][1] in has_choose:
iou = similarity.iou(matching_pair[pair_index][2], ans_gt[matching_pair[pair_index][0]])
if iou >= 0.5:
has_choose.append(matching_pair[pair_index][1])
ans_list.remove(matching_pair[pair_index][0])
ans_hit+= float(1) / len(ans_gt)/ans_length[matching_pair[pair_index][0]]
else:
ans_length[matching_pair[pair_index][0]]+= 1
if not ans_list:
break
# new start
img_ans_list = []
top_k = 100
for start_index in xrange(top_k):
has_choose = []
ans_list = [i for i in xrange(len(ans_gt))]
img_ans = []
img_ans_score = 0
for pair_index in xrange(start_index, len(matching_pair)):
if matching_pair[pair_index][0] in ans_list and not matching_pair[pair_index][1] in has_choose:
has_choose.append(matching_pair[pair_index][1])
ans_list.remove(matching_pair[pair_index][0])
img_ans_score += matching_pair[pair_index][3]
img_ans.append([matching_pair[pair_index][0], matching_pair[pair_index][2]])
if not ans_list:
img_ans_list.append([img_ans, img_ans_score])
break
img_ans_list.sort(key = lambda x: float(x[1]), reverse = True)
after_dcg = []
for img_ans in img_ans_list:
img_ans, score = img_ans
gt_score = []
for single in img_ans:
[gt, gps_pos, center_pos, yaw] = gt_list[single[0]]
bb = single[1]
p1 = Point(gps_pos[1], gps_pos[0])
origin_p = Point(center_pos[1], center_pos[0])
gt_score.append([float(bb[2]+bb[0])/2
, similarity.disofpoint2line(p1, origin_p, similarity.yawtovector(yaw))])
gt_score.sort(key = lambda x: float(x[1]), reverse = True)
dcg_list = [gs[0] for gs in gt_score]
after_dcg.append([img_ans, score+10*similarity.sigmoid(similarity.DCG(dcg_list)/5000)])
after_dcg.sort(key = lambda x: float(x[1]), reverse = True)
img_ans = after_dcg[0][0]
for single in img_ans:
#IPython.embed()
if similarity.iou(single[1], ans_gt[single[0]]) > 0.5:
img_sum.append(1)
else:
img_sum.append(0)
"""
#IPython.embed()
#print("-------------------------------------")
#print(rr)
#print(hit)
#print(poi_num)
print("{:.4f}".format(hit/poi_num))
print("Whole: {:.4f}".format(ans_hit/img_num))
#print("multi: {:.4f}".format(np.mean(img_sum)))
print("# of Img: {}".format(img_num))
print("# of POI: {}".format(poi_num))
print("# Never : {}".format(never))
#embed()
#145 383
if __name__ == '__main__':
main()
"""
https://www.google.com.tw/maps/place/25%C2%B001'38.5%22N+121%C2%B032'09.2%22E/@25.0287444,121.5360478,15z/data = !4m5!3m4!1s0x0:0x0!8m2!3d25.027352!4d121.535887
https://maps.googleapis.com/maps/api/place/nearbysearch/json?location=25.0415889,121.5640499&radius=2000&key=AIzaSyDAnTTBRbpMAinjfngLWKQWVW1w-OojBLE&type=establishment&keyword=%E8%87%BA%E5%8C%97%E5%B0%8F%E5%B7%A8%E8%9B%8B
https://maps.googleapis.com/maps/api/streetview?size=600x600&location=25.037525,121.563782&pitch=30&key=AIzaSyDAnTTBRbpMAinjfngLWKQWVW1w-OojBLE
http://www.cmlab.csie.ntu.edu.tw/~jacky82226/demo/retrieval.html
"""