-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtesting.py
480 lines (443 loc) · 29.8 KB
/
testing.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
import _init_paths
import caffe
import tools
import os.path as osp
import numpy as np
from caffe import layers as L, params as P, to_proto
import matplotlib.pyplot as plt
from PIL import Image
import global_var as GV
import matplotlib.image as mping
import h5py
import scipy.io as scio
import os
from pdb import set_trace
#das
def BN_scale_relu_conv(split, bottom, nout , ks=3, stride=1, pad=1, dilation = 1, conv_type = 'conv', in_place = False, lr = 2):
if dilation != 1 and conv_type == 'conv':
pad = ((dilation - 1) * (ks - 1) + ks - 1) / 2
# ks = (dilation - 1) * (ks - 1) + ks
# dilation = 1
# dsa
if split == 'train':
BN = L.BatchNorm(bottom, batch_norm_param = dict(use_global_stats = False), in_place=in_place,
param = [dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0)])
else:
BN = L.BatchNorm(bottom, batch_norm_param = dict(use_global_stats = True), in_place=in_place,
param = [dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0)])
scale = L.Scale(BN, scale_param = dict(bias_term = True), in_place = True)
relu = L.ReLU(scale, in_place=True)
if conv_type == 'conv':
conv = L.Convolution(relu, kernel_size=ks, stride=stride, dilation = dilation, num_output=nout, pad=pad, bias_term = True, #, std = 0.000000001, mean = 0
weight_filler = dict(type='xavier'),
bias_filler = dict(type='constant'),
param=[dict(lr_mult=lr/2, decay_mult=1), dict( lr_mult=lr, decay_mult=0)])
elif conv_type == 'dconv':
conv = L.Deconvolution(relu, convolution_param=dict( weight_filler = dict(type='xavier'), dilation = dilation, num_output = nout, kernel_size = ks, stride = stride, pad = pad, bias_term=False), param=[dict(lr_mult=1)])
return BN, scale, relu, conv
def conv_BN_scale_relu(split, bottom, nout , ks=3, stride=1, pad=1, dilation = 1, conv_type = 'conv', in_place = True, lr = 2):
if dilation != 1 and conv_type == 'conv':
pad = ((dilation - 1) * (ks - 1) + ks - 1) / 2
# ks = (dilation - 1) * (ks - 1) + ks
# dilation = 1
if conv_type == 'conv':
conv = L.Convolution(bottom, kernel_size=ks, stride=stride, dilation = dilation, num_output=nout, pad=pad, bias_term = True, #, std = 0.000000001, mean = 0
weight_filler = dict(type='xavier'),
bias_filler = dict(type='constant'),
param=[dict(lr_mult=lr/2, decay_mult=1), dict(lr_mult=lr, decay_mult=0)])
elif conv_type == 'dconv':
conv = L.Deconvolution(bottom, convolution_param=dict( weight_filler = dict(type='xavier'), dilation = dilation, num_output = nout, kernel_size = ks, stride = stride, pad = pad, bias_term=False), param=[dict(lr_mult=1)])
if split == 'train':
BN = L.BatchNorm(conv, batch_norm_param = dict(use_global_stats = False), in_place=in_place,
param = [dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0)])
else:
BN = L.BatchNorm(conv, batch_norm_param = dict(use_global_stats = True), in_place=in_place,
param = [dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0),dict(lr_mult=0, decay_mult=0)])
scale = L.Scale(BN, scale_param = dict(bias_term = True), in_place = in_place)
relu = L.ReLU(scale, in_place=in_place)
return conv, BN, scale, relu
def max_pool(bottom, ks=2, stride=2):
return L.Pooling(bottom, pool=P.Pooling.MAX, kernel_size=ks, stride=stride)
test_num = 3
# ＨＥＲＥ
GV.test_data_dir = 'test_data/Images_test' # '0322data/Train-n/Images_train' 0305_data_split 0331data/Train-n/Images_train
GV.test_labels_dir = 'test_data/test_mask'
GV.test_sobel_dir = 'test_data/Images_test'
GV.lr = 0.1
def simpleNN(split):
nout = 32
repeat = 2
dilation = 1
ks = 3
pad = 1
block_nums = 1
block_nums = block_nums - 1
GV.repeat = repeat
data, labels = L.Python(module='read_data1',
layer='input_layer',
ntop=2,
param_str=str(dict(split=split,
# data_dir = './data/../../',
data_dir='/content/drive/My Drive/MEnet',
train_data_dir='0408data1/Images_train',
train_sobel_dir='0408data1/Images_train_SOBEL',
train_labels_dir='0408data1/GT_train',
test_data_dir=GV.test_data_dir, #
test_sobel_dir=GV.test_sobel_dir,
# test_sobel_dir='0408data1/Images_train_SOBEL',
test_labels_dir=GV.test_labels_dir, #
batch_size=5,
test_batch_size=1,
resize_size=[224, 224]
)))
# gf0 = L.Python(data, module = 'filter',
# layer = 'guided_filer',
# ntop = 1,
# param_str = str(dict(split=split,
# radius = 15,
# epsilon = 0.3,
# guided_type = 'gray'
# )))
set_dilation = 1
dilation = set_dilation
# gf0 = L.Concat(gf0, sobel_0)
result = data
a, b, c, result = conv_BN_scale_relu(split, bottom = result, nout = nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# data = L.Concat(data, sobel_0)
a, b, c, data_f = BN_scale_relu_conv(split, bottom = data, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, mask_f = BN_scale_relu_conv(split, bottom = data, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# data_f = L.Eltwise(data_f, mask_f, operation = P.Eltwise.PROD)
for i in range(repeat): #224
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
scale0 = result
# scale0 = L.Concat(scale0, sobel_0)
a, b, c, scale0_m = BN_scale_relu_conv(split, bottom = scale0, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale0_d_mask = BN_scale_relu_conv(split, bottom = scale0, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale0_m = L.Eltwise(scale0_m, scale0_d_mask, operation = P.Eltwise.PROD)
for i in range(repeat):#112
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 2 * nout, ks=ks, stride=2, pad = pad, dilation = dilation, in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 2 * nout, ks=1, stride=2, pad = 0, dilation = dilation, in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 2 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 2 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
scale1 = result
# scale1 = L.Concat(scale1, sobel_1)
a, b, c, scale1_m = BN_scale_relu_conv(split, bottom = scale1, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale1_d_mask = BN_scale_relu_conv(split, bottom = scale1, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale1_m = L.Eltwise(scale1_m, scale1_d_mask, operation = P.Eltwise.PROD)
for i in range(repeat):#56
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 4 * nout, ks=ks, stride=2, pad = pad, dilation = dilation, in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 4 * nout, ks=1, stride=2, pad = 0, dilation = dilation, in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 4 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 4 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
scale2 = result
# scale2 = L.Concat(scale2, sobel_2)
a, b, c, scale2_m = BN_scale_relu_conv(split, bottom = scale2, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale2_d_mask = BN_scale_relu_conv(split, bottom = scale2, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale2_m = L.Eltwise(scale2_m, scale2_d_mask, operation = P.Eltwise.PROD)
for i in range(repeat):#28
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 8 * nout, ks=ks, stride=2, pad = pad, dilation = dilation, in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 8 * nout, ks=1, stride=2, pad = 0, dilation = dilation, in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 8 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 8 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
scale3 = result
# scale3 = L.Concat(scale3, sobel_3)
a, b, c, scale3_m = BN_scale_relu_conv(split, bottom = scale3, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale3_d_mask = BN_scale_relu_conv(split, bottom = scale3, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale3_m = L.Eltwise(scale3_m, scale3_d_mask, operation = P.Eltwise.PROD)
for i in range(repeat):#14
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 16 * nout, ks=ks, stride=2, pad = pad, dilation = dilation, in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 16 * nout, ks=1, stride=2, pad = 0, dilation = dilation, in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 16 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 16 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
scale4 = result
# scale4 = L.Concat(scale4, sobel_4)
a, b, c, scale4_m = BN_scale_relu_conv(split, bottom = scale4, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale4_d_mask = BN_scale_relu_conv(split, bottom = scale4, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale4_m = L.Eltwise(scale4_m, scale4_d_mask, operation = P.Eltwise.PROD)
for i in range(repeat):#7
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 32 * nout, ks=ks, stride=2, pad = pad, dilation = dilation, in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 32 * nout, ks=1, stride=2, pad = 0, dilation = dilation, in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 32 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 32 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
scale5 = result
# scale5 = L.Concat(scale5, sobel_5)
a, b, c, scale5_m = BN_scale_relu_conv(split, bottom = scale5, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale5_d_mask = BN_scale_relu_conv(split, bottom = scale5, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale5_m = L.Eltwise(scale5_m, scale5_d_mask, operation = P.Eltwise.PROD)
dilation = 1
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 32 * nout, ks=7, stride=1, pad = 0, dilation = dilation, in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 32 * nout, ks =7, stride=1, pad = 0, dilation = dilation, conv_type = 'dconv', in_place = False)
a, b, c, scale5_u = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale5_u_mask = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale5_u = L.Eltwise(scale5_u, scale5_u_mask, operation = P.Eltwise.PROD)
result = L.Concat(result, scale5)
for i in range(repeat):#14
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 16 * nout, ks = 4, stride=2, pad = 1, dilation = dilation, conv_type = 'dconv', in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 16 * nout, ks = 2, stride=2, pad = 0, dilation = dilation, conv_type = 'dconv', in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 16 * nout, ks=ks, stride=1, pad = pad, dilation = dilation,in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 16 * nout, ks=ks, stride=1, pad = pad, dilation = dilation ,in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
a, b, c, scale4_u = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale4_u_mask = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale4_u = L.Eltwise(scale4_u, scale4_u_mask, operation = P.Eltwise.PROD)
result = L.Concat(result, scale4)
for i in range(repeat):#28
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 8 * nout, ks = 4, stride=2, pad = 1, dilation = dilation, conv_type = 'dconv', in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 8 * nout, ks = 2, stride=2, pad = 0, dilation = dilation, conv_type = 'dconv', in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 8 * nout, ks=ks, stride=1, pad = pad, dilation = dilation,in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 8 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
a, b, c, scale3_u = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale3_u_mask = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale3_u = L.Eltwise(scale3_u, scale3_u_mask, operation = P.Eltwise.PROD)
result = L.Concat(result, scale3)
for i in range(repeat):#56
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 4 * nout, ks = 4, stride=2, pad = 1, dilation = dilation, conv_type = 'dconv', in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 4 * nout, ks = 2, stride=2, pad = 0, dilation = dilation, conv_type = 'dconv', in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 4 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 4 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
a, b, c, scale2_u = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale2_u_mask = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale2_u = L.Eltwise(scale2_u, scale2_u_mask, operation = P.Eltwise.PROD)
result = L.Concat(result, scale2)
for i in range(repeat):#112
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 2 * nout, ks = 4, stride=2, pad = 1, dilation = dilation, conv_type = 'dconv', in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 2 * nout, ks = 2, stride=2, pad = 0, dilation = dilation, conv_type = 'dconv', in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 2 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 2 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
a, b, c, scale1_u = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale1_u_mask = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale1_u = L.Eltwise(scale1_u, scale1_u_mask, operation = P.Eltwise.PROD)
result = L.Concat(result, scale1)
for i in range(repeat):#224
if i == 0:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 1 * nout, ks = 4, stride=2, pad = 1, dilation = dilation, conv_type = 'dconv', in_place = False)
a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 1 * nout, ks = 2, stride=2, pad = 0, dilation = dilation, conv_type = 'dconv', in_place = False)
else:
a, b, c, result1 = BN_scale_relu_conv(split, bottom = result, nout = 1 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
a, b, c, result2 = BN_scale_relu_conv(split, bottom = result1, nout = 1 * nout, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = True)
result = L.Eltwise(result2, result, operation = P.Eltwise.SUM)
a, b, c, scale0_u = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# a, b, c, scale0_u_mask = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
# scale0_u = L.Eltwise(scale0_u, scale0_u_mask, operation = P.Eltwise.PROD)
# result = L.Concat(result, scale0)
# a, b, c, result = BN_scale_relu_conv(split, bottom = result, nout = 1, ks=ks, stride=1, pad = pad, dilation = dilation, in_place = False)
result = L.Python(data_f, scale0_m, scale1_m, scale2_m, scale3_m, scale4_m, scale5_m, scale0_u, scale1_u, scale2_u, scale3_u, scale4_u, scale5_u,
ntop = 1,
module = 'resize',
layer = 'resize_to_same_size',
param_str = str(dict()))
# metric = result
a, b, metric, c = conv_BN_scale_relu(split, bottom = result, nout = 16, ks=3, stride=1, pad = 1, dilation = dilation, in_place = True)
metric = L.TanH(metric, in_place= True)
a, b, sal, d = conv_BN_scale_relu(split, bottom = result, nout = 2, ks=3, stride=1, pad = 1, dilation = dilation, in_place = True)
result = L.Python(metric, sal, labels,
ntop = 1,
module = 'pyloss',
layer = 'EuclideanLossLayer',
param_str = str(dict(crop_size = [64, 64],
batches = 20,
sup_threshold = 3,
inf_threshold = 0
)),loss_weight=1)
return to_proto(result)
def store_curve_value(snapshot_name, train_dir, test_dir, data_iter = 10, iter_step = 500):
loss_file = snapshot_name + '.h5'
start_index = 0
if os.path.exists(loss_file):
print('The file ' + loss_file + '.h5 exists.' )
h5file = h5py.File(loss_file, 'r')
train_metric_loss = h5file['train_metric_loss'][...]
train_sal_loss = h5file['train_sal_loss'][...]
test_metric_loss = h5file['test_metric_loss'][...]
test_sal_loss = h5file['test_sal_loss'][...]
h5file.close()
start_index = (train_metric_loss.shape[0]) * iter_step
train_metric_loss = list(train_metric_loss)
train_sal_loss = list(train_sal_loss)
test_metric_loss = list(test_metric_loss)
test_sal_loss = list(test_sal_loss)
print(start_index)
else:
train_metric_loss = []
train_sal_loss = []
test_metric_loss = []
test_sal_loss = []
print('The file ' + loss_file + ' has set up.')
for index in range(start_index + iter_step, 110000 + iter_step, iter_step):
# print index
weight = osp.join('model', 'snapshot', snapshot_name + '_iter_' + str(index) + '.caffemodel')
train_net = caffe.Net(str(train_dir), str(weight), caffe.TEST)
test_net = caffe.Net(str(test_dir), str(weight), caffe.TEST)
tmp_train_metric_loss = 0
tmp_train_sal_loss = 0
tmp_test_metric_loss = 0
tmp_test_sal_loss = 0
for forward_iter in range(data_iter):
train_net.forward()
test_net.forward()
tmp_train_metric_loss = tmp_train_metric_loss + train_net.blobs['Python4'].data[0]
tmp_train_sal_loss = tmp_train_sal_loss + train_net.blobs['Python4'].data[1]
tmp_test_metric_loss = tmp_test_metric_loss + test_net.blobs['Python4'].data[0]
tmp_test_sal_loss = tmp_test_sal_loss + test_net.blobs['Python4'].data[1]
train_metric_loss.append(train_metric_loss / data_iter)
train_sal_loss.append(train_sal_loss / data_iter)
test_metric_loss.append(test_metric_loss / data_iter)
test_sal_loss.append(test_sal_loss / data_iter)
# if index % 20 == 0:
print('saving loss data', train_metric_loss, train_sal_loss, test_metric_loss, test_sal_loss)
# da
h5file = h5py.File(loss_file, 'w')
h5file.create_dataset('train_metric_loss', data = np.array(train_metric_loss), dtype = np.float32)
h5file.create_dataset('train_sal_loss', data = np.array(train_sal_loss), dtype = np.float32)
h5file.create_dataset('test_metric_loss', data = np.array(test_metric_loss), dtype = np.float32)
h5file.create_dataset('test_sal_loss', data = np.array(test_sal_loss), dtype = np.float32)
h5file.close()
# store_curve_value(snapshot_name, train_dir, test_dir)
# das
def store_curve_to_mat(snapshot_name, train_dir, test_dir, data_iter = 100, iter_step = 10000):
snapshot_name = snapshot_name.split('_iter_')[0]
loss_file = snapshot_name + '.mat'
start_index = 0
if os.path.exists(loss_file):
print('The file ' + loss_file + '.mat exists.')
matfile = scio.loadmat(loss_file)
train_metric_loss = matfile['train_metric_loss']
train_sal_loss = matfile['train_sal_loss']
test_metric_loss = matfile['test_metric_loss']
test_sal_loss = matfile['test_sal_loss']
start_index = (train_metric_loss.shape[0]) * iter_step
train_metric_loss = list(train_metric_loss)
train_sal_loss = list(train_sal_loss)
test_metric_loss = list(test_metric_loss)
test_sal_loss = list(test_sal_loss)
print(start_index)
else:
train_metric_loss = []
train_sal_loss = []
test_metric_loss = []
test_sal_loss = []
print('The file ' + loss_file + ' has set up.')
for index in range(start_index + iter_step, 110000 + iter_step, iter_step):
# print index
weight = osp.join('model', 'snapshot', snapshot_name + '_iter_' + str(index) + '.caffemodel')
train_net = caffe.Net(str(train_dir), str(weight), caffe.TEST)
test_net = caffe.Net(str(test_dir), str(weight), caffe.TEST)
tmp_train_metric_loss = 0
tmp_train_sal_loss = 0
tmp_test_metric_loss = 0
tmp_test_sal_loss = 0
for forward_iter in range(data_iter):
train_net.forward()
test_net.forward()
tmp_train_metric_loss = tmp_train_metric_loss + train_net.blobs['Python4'].data[0]
tmp_train_sal_loss = tmp_train_sal_loss + train_net.blobs['Python4'].data[1]
tmp_test_metric_loss = tmp_test_metric_loss + test_net.blobs['Python4'].data[0]
tmp_test_sal_loss = tmp_test_sal_loss + test_net.blobs['Python4'].data[1]
train_metric_loss.append(tmp_train_metric_loss / data_iter)
train_sal_loss.append(tmp_train_sal_loss / data_iter)
test_metric_loss.append(tmp_test_metric_loss / data_iter)
test_sal_loss.append(tmp_test_sal_loss / data_iter)
# if index % 20 == 0:
print('saving loss data', train_metric_loss, train_sal_loss, test_metric_loss, test_sal_loss)
# da
scio.savemat(loss_file, {'train_metric_loss': train_metric_loss,
'train_sal_loss': train_sal_loss,
'test_metric_loss': test_metric_loss,
'test_sal_loss': test_sal_loss})
# h5file = h5py.File(loss_file, 'w')
# h5file.create_dataset('train_metric_loss', data = np.array(train_metric_loss), dtype = np.float32)
# h5file.create_dataset('train_sal_loss', data = np.array(train_sal_loss), dtype = np.float32)
# h5file.create_dataset('test_metric_loss', data = np.array(test_metric_loss), dtype = np.float32)
# h5file.create_dataset('test_sal_loss', data = np.array(test_sal_loss), dtype = np.float32)
# h5file.close()
def make_net(snapshot_name = None):
# this_dir + '/model/train.prototxt'
if not snapshot_name:
train_prototxt_dir = './model/train.prototxt'
test_prototxt_dir = './model/test.prototxt'
else:
train_prototxt_dir = './model/train_{}.prototxt'.format(snapshot_name)
test_prototxt_dir = './model/test_{}.prototxt'.format(snapshot_name)
with open(train_prototxt_dir, 'w') as f:
f.write(str(simpleNN('train')))
with open(test_prototxt_dir, 'w') as f:
f.write(str(simpleNN('test')))
solver_dir = './model/solver.prototxt'
return train_prototxt_dir, test_prototxt_dir, solver_dir
GV.hard_samples = []
this_dir = osp.dirname(__file__)
GV.test_nums = 0
GV.test_images_num = 3
GV.abnormal_files = []
watch_single_result = True
#GV.phase = 'train'
GV.phase = 'test'
GV.test_dir = 'test_data/Images_test'
GV.target_data_dir = '.'
GV.target_dir = GV.target_data_dir
if __name__ == '__main__':
# make_net()
#GV.phase = 'test'
'''
f4_lr_0.1_0408_pyloss_data_sal_split_first5times_pyloss_plus_num_finalout16_only_sal_Loss_ks3_HardNeg_crop_nout32_repeat2_batch5_dilation1_1e-8_iter_110000
f4_lr_0.1_0408_pyloss_data_sal_split_first5times_pyloss_plus_num_finalout16_only_metric_Loss_ks3_HardNeg_crop_nout32_repeat2_batch5_dilation1_1e-8_iter_110000
f4_lr_0.1_0408_pyloss_data_sal_split_first5times_pyloss_plus_num_finalout16_two_Loss_ks3_HardNeg_crop_nout32_repeat2_batch5_dilation1_1e-8_iter_110000
'''
#train_dir, test_dir, solver_dir = make_net()
caffe.set_device(2)
caffe.set_mode_gpu()
#snapshot_name = 'f4_lr_0.1_0408_pyloss_data_sal_split_first5times_pyloss_plus_num_finalout16_two_Loss_ks3_HardNeg_crop_nout32_repeat2_batch5_dilation1_1e-8_iter_110000'
snapshot_name = 'attention_iter_110000'
weight = '/workspace/model/snapshot/' + str(snapshot_name) + '.caffemodel'
state = '/workspace/model/snapshot/' + str(snapshot_name) + '.solverstate'
# if GV.phase == 'test':
# net = caffe.Net(str(test_dir), str(weight), caffe.TEST)
# net.forward()
#set_trace()
if GV.phase == 'test':
test_files = os.listdir(osp.join(GV.test_dir))
# print('1:',test_files)
GV.target_data_dir = snapshot_name
train_dir, test_dir, solver_dir = make_net(snapshot_name = snapshot_name)
set_trace()
net = caffe.Net(str(test_dir), str(weight), caffe.TEST)
# dsd
# store_curve_to_mat(snapshot_name, train_dir, test_dir)
net.forward()
for p in range(GV.test_images_num):
print('now is Number:',p)
net.forward()
elif GV.phase == 'train':
solver = caffe.SGDSolver(str(solver_dir))
solver.step(1)
for i in range(1100):
solver.step(100)
print('done!')