-
Notifications
You must be signed in to change notification settings - Fork 2
/
_test.py
350 lines (274 loc) · 13.2 KB
/
_test.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
import numpy as np
import cv2
import os
import glob
from scipy import signal
# DL-methods
from model.PNN import PNNModel
from model.PanNet import PanNetModel
from model.Ours import OursModel
from model.TFNet import TFNetModel
from model.MSDCNN import MSDCNNModel
from model.SRPPNN import SRPPNNModel
import cupy as cp
from cupy.core.dlpack import toDlpack
from cupy.core.dlpack import fromDlpack
from torch.utils.dlpack import to_dlpack
from torch.utils.dlpack import from_dlpack
import gdal
from tqdm import tqdm
import torch
import time
from thop import profile
import argparse
from mlab.releases import latest_release as matlab
def ERGAS(I1, I2, ratio=4):
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.ERGAS(I1,I2,[ratio])#调用自己定义的m函数就可以了
return index
def SAM(I1,I2):
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.SAM(I1,I2)#调用自己定义的m函数就可以了
return index
def SCC(I1,I2):
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.SCC(I1,I2)#调用自己定义的m函数就可以了
return index
def Q(I1,I2):
# I1 融合
# I2 GT
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.q2n(I2,I1, [32], [32])#调用自己定义的m函数就可以了
return index
def QAVE(I1,I2):
# I1 融合
# I2 GT
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.Q(I2,I1, [2048])#调用自己定义的m函数就可以了
return index
def D_lambda(I1,I2):
# I1 融合
# I2 放大MS
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.D_lambda(I1,I2, [32], [1])#调用自己定义的m函数就可以了
return index
def D_s(I1,I2,I3, args):
# I1 融合
# I2 放大MS
# I3 PAN
matlab.path(matlab.path(),r'D:\实验\影像融合\其他融合方法\Pansharpening Tool ver 1.3\Quality_Indices')#设置路径
index = matlab.D_s(I1,I2,I3, str(args.sensor),[args.scale],[32], [1])#调用自己定义的m函数就可以了
return index
def QNR(D_s, D_lambda):
alpha = 1
beta = 1
return (1-D_lambda)**alpha * (1-D_s)**beta
def gdal_read(input_file):
# 读取多通道
# the output format is (h,w,c)
dataset = gdal.Open(input_file)
geo = dataset.GetGeoTransform()
proj = dataset.GetProjection()
data = dataset.ReadAsArray() # in order to fit the torch.fromnumpy
if len(data.shape) == 2: # this is one band
return geo, proj, data.reshape(data.shape[0], data.shape[1], 1)
else:
return geo, proj, data.transpose((1,2,0)) # in order to use the cv transform
def gdal_write(output_file,array_data, geo, proj):
#判断栅格数据的数据类型0
if 'int8' in array_data.dtype.name:
datatype = gdal.GDT_Byte
elif 'int16' in array_data.dtype.name:
datatype = gdal.GDT_UInt16
else:
datatype = gdal.GDT_Float32
h,w,c = array_data.shape
Driver = gdal.GetDriverByName("Gtiff")
dataset = Driver.Create(output_file,w,h,c,datatype)
dataset.SetGeoTransform(geo)
dataset.SetProjection(proj)
for i in range(c):
band = dataset.GetRasterBand(i+1)
band.WriteArray(array_data[:,:,i])
parser = argparse.ArgumentParser()
parser.add_argument("-m", "--model", type=str, default="ours", help="deep learning pan-sharpening method such as PNN MSDCNN PanNet TFNet SRPPNN and DIPNet")
parser.add_argument("-ms", "--model_sub", type=str, default="", help="for our DIPNet's ablation study, such as SR SR_PLB SR_PHB SR_PLB_PHB default set as ''")
parser.add_argument("-ml", "--model_loss", type=str, default="", help="for our DIPNet's ablation study, such as SSIM or L1 or '', default set as ''")
parser.add_argument("-t", "--isTrain", type=bool, default=True, help="whether or not to train")
parser.add_argument("-g", "--gpu_ids", type=int, default=0, help="")
parser.add_argument("-c", "--continue_train", type=bool, default=False, help="whether or not to continue")
parser.add_argument("-we", "--which_epoch", type=int, default=-1, help="if continue Train ture, set this value")
parser.add_argument("-p", "--print_net_in_detail", type=bool, default=False, help="whether or not to continue")
parser.add_argument("-d", "--dataDir", type=str, default="", help="this is a train dataset dir")
parser.add_argument("-td", "--testdataDir", type=str, default="", help="this is a test dataset dir ")
parser.add_argument("-sr", "--save_result", type=bool, default=True, help="whether or not to save result")
parser.add_argument("-sd", "--saveDir", type=str, default="", help="save directory")
parser.add_argument("-cd", "--checkpoints_dir", type=str, default="", help="the dir of pertrained models or the dir to checkpoint the model parameters during training")
parser.add_argument("-nT", "--nThreads", type=int, default=0, help="use serval threads to load data in pytorch")
parser.add_argument("-bs", "--batchSize", type=int, default=16, help="")
parser.add_argument("-is", "--img_size", type=int, default=32, help="simulate ms size")
parser.add_argument("-sa", "--scale", type=int, default=4, help="scale factor which resizing to pan size")
parser.add_argument("-se", "--seed", type=int, default=19970716, help="random seed")
parser.add_argument("-pf", "--print_freq", type=int, default=5, help="print frequency of log")
parser.add_argument("-sf", "--save_epoch_freq", type=int, default=500, help="print frequency of log")
parser.add_argument("-pc", "--pan_channel", type=int, default=1, help="pan-chromatic band")
parser.add_argument("-mc", "--mul_channel", type=int, default=4, help="multi-spectral band which is based on different satellite")
parser.add_argument("-gm", "--gan_mode", type=str, default="lsgan", help="'lsgan' or 'wgangp' or 'vanilla' this is orginal")
parser.add_argument("-dr", "--data_range", type=int, default=2047, help="radis resolution ")
parser.add_argument("-lp", "--lr_policy", type=str, default='step', help="")
parser.add_argument("-ot", "--optim_type", type=str, default='adam', help="optim_type")
parser.add_argument("-lr", "--lr", type=float, default=1e-4, help="learning rate")
parser.add_argument("-b", "--beta", type=float, default=0.9, help="")
parser.add_argument("-mo", "--momentum", type=float, default=0.9, help="")
parser.add_argument("-w", "--weight_decay", type=float, default=1e-8, help="")
parser.add_argument("-li", "--lr_decay_iters", type=list, default=[1000], help="")
parser.add_argument("-lf", "--lr_decay_factor", type=float, default=0.5, help="")
parser.add_argument("-e", "--epochs", type=int, default=1000, help="")
parser.add_argument("-iu", "--isUnlabel", type=bool, default=False, help="")
parser.add_argument("-ie", "--isEval", type=bool, default=False, help="")
parser.add_argument("-ia", "--isAug", type=bool, default=True, help="")
parser.add_argument("-uf", "--useFakePAN", type=bool, default=False, help="")
parser.add_argument("-sensor", "--sensor", type=str, default="", help="")
args = parser.parse_args()
input_data_path = args.testdataDir
input_ms_path = glob.glob(os.path.join(os.path.join(input_data_path, 'MS'), '*.tif'))
input_pan_path = glob.glob(os.path.join(os.path.join(input_data_path, 'PAN'), '*.tif'))
print(input_ms_path, input_pan_path)
input_ms_down_path = glob.glob(os.path.join(os.path.join(input_data_path, 'LR'), '*.tif'))
input_pan_down_path = glob.glob(os.path.join(os.path.join(input_data_path, 'HR'), '*.tif'))
# 创建存储路径
if args.save_result:
# dl方法
ours_path = os.path.join(args.saveDir, args.model+args.model_sub+args.model_loss)
# dl supervised
supervised_path = os.path.join(ours_path, 'supervised')
reduce_path = os.path.join(supervised_path, 'reduce')
full_path = os.path.join(supervised_path, 'full')
if not os.path.exists(ours_path):
os.makedirs(ours_path)
if not os.path.exists(supervised_path):
os.makedirs(supervised_path)
if not os.path.exists(reduce_path):
os.makedirs(reduce_path)
if not os.path.exists(full_path):
os.makedirs(full_path)
# dl unsupervised
ours_path = os.path.join(args.saveDir, args.model)
unsupervised_path = os.path.join(ours_path, 'unsupervised')
fake_path = os.path.join(unsupervised_path, 'fake')
unfull_path = os.path.join(unsupervised_path, 'full')
unreduce_path = os.path.join(unsupervised_path, 'reduce')
if not os.path.exists(unsupervised_path):
os.makedirs(unsupervised_path)
if not os.path.exists(fake_path):
os.makedirs(fake_path)
if not os.path.exists(unfull_path):
os.makedirs(unfull_path)
if not os.path.exists(unreduce_path):
os.makedirs(unreduce_path)
# 选择
if args.model == 'ours':
model = OursModel()
elif args.model == 'PNN':
model = PNNModel()
elif args.model == 'PanNet':
model = PanNetModel()
elif args.model == 'TFNet':
model = TFNetModel()
elif args.model == 'MSDCNN':
model = MSDCNNModel()
elif args.model == 'SRPPNN':
model = SRPPNNModel()
model.initialize(args)
if args.scale == 6:
input1 = torch.randn(1, args.mul_channel, 64, 64).cuda()
input2 = torch.randn(1, args.pan_channel, 384, 384).cuda()
else:
input1 = torch.randn(1, args.mul_channel, 64, 64).cuda()
input2 = torch.randn(1, args.pan_channel, 256, 256).cuda()
flop, para = profile(model.netG, inputs=(input1, input2, ))
print("%.2fM" % (flop/1e6), "%.2fM" % (para/1e6))
print('Total params: %.2fM' % (sum(p.numel() for p in model.netG.parameters())/1000000.0))
model.load_networks(999)
full_times = []
reduce_times = []
ergass = []
sams = []
sccs = []
qs = []
qaves = []
dss = []
d_ls = []
qnrs = []
for i, ms_path in tqdm(enumerate(input_ms_path)):
# print(i)
step = i
'''loading data'''
_, _, used_ms = gdal_read(ms_path)
geo, proj, used_pan = gdal_read(input_pan_path[i])
_, _, used_ms_down = gdal_read(input_ms_down_path[i])
geo_down, proj_down, used_pan_down = gdal_read(input_pan_down_path[i])
'''normalization'''
used_ms = used_ms / args.data_range
used_pan = used_pan / args.data_range
used_ms_down = used_ms_down / args.data_range
used_pan_down = used_pan_down / args.data_range
name = os.path.basename(ms_path)
if args.isUnlabel:
fused_image = model.predict(used_ms[:, :, :], used_pan[:, :, :])
if args.save_result:
gdal_write(os.path.join(unfull_path,name), fused_image, geo)
fused_down_image = model.predict(used_ms_down[:, :, :], used_pan_down[:, :, :])
if args.save_result:
gdal_write(os.path.join(unreduce_path,name), fused_down_image, geo_down)
else:
s_time = time.time()
fused_image = model.predict(used_ms[:, :, :], used_pan[:, :, :])
e_time = time.time()
full_times.append(e_time-s_time)
# ds = D_s(fused_image, cv2.resize(used_ms, dsize=(0,0),fx=args.scale, fy=args.scale), used_pan, args)
# d_l = D_lambda(fused_image, cv2.resize(used_ms, dsize=(0,0),fx=args.scale, fy=args.scale))
# qnr = QNR(ds, d_l)
# dss.append(ds)
# d_ls.append(d_l)
# qnrs.append(qnr)
if args.save_result:
gdal_write(os.path.join(full_path,name), fused_image, geo, proj)
s_time = time.time()
fused_down_image = model.predict(used_ms_down[:, :, :], used_pan_down[:, :, :])
e_time = time.time()
reduce_times.append(e_time-s_time)
# ergas = ERGAS(fused_down_image, used_ms)
# sam = SAM(fused_down_image, used_ms)
# scc = SCC(fused_down_image, used_ms)
# q = Q(fused_down_image, used_ms)
# qave = QAVE(fused_down_image, used_ms)
# ergass.append(ergas)
# sams.append(sam)
# sccs.append(scc)
# qs.append(q)
# qaves.append(qave)
if args.save_result:
gdal_write(os.path.join(reduce_path,name), fused_down_image, geo_down, proj_down)
full_times = np.array(full_times)
reduce_times = np.array(reduce_times)
ergass = np.array(ergass)
sams = np.array(sams)
sccs = np.array(sccs)
qs = np.array(qs)
qaves = np.array(qaves)
dss = np.array(dss)
d_ls = np.array(d_ls)
qnrs = np.array(qnrs)
print(args.sensor)
print(args.model+args.model_sub+args.model_loss)
print('full_time:{}±{}'.format(np.mean(full_times), np.var(full_times)))
print('reduce_time:{}±{}'.format(np.mean(reduce_times), np.var(reduce_times)))
print('ergas:{}±{}'.format(np.mean(ergass), np.var(ergass)))
print('sam:{}±{}'.format(np.mean(sams), np.var(sams)))
print('scc:{}±{}'.format(np.mean(sccs), np.var(sccs)))
print('q:{}±{}'.format(np.mean(qs), np.var(qs)))
print('qave:{}±{}'.format(np.mean(qaves), np.var(qaves)))
print('ds:{}±{}'.format(np.mean(dss), np.var(dss)))
print('d_l:{}±{}'.format(np.mean(d_ls), np.var(d_ls)))
print('qnr:{}±{}'.format(np.mean(qnrs), np.var(qnrs)))