-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathv8_test.py
188 lines (168 loc) · 6.71 KB
/
v8_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
# -*- coding: utf-8 -*-
"""
author: lhl
"""
import torch
import os
import numpy as np
import time
import random
import gdal
from v8_train import StfNet, Loss
from v8_train import MyDataset
import utilme
import argparse
import torch.optim as optim
from torch import nn, einsum
from einops import rearrange, repeat
from einops.layers.torch import Rearrange
from itertools import chain
from tqdm.notebook import tqdm
# from torch.utils.tensorboard import SummaryWriter
from torch.utils.data import DataLoader, Dataset
from torch.optim import lr_scheduler
from timm.scheduler.cosine_lr import CosineLRScheduler
# def seed_everything(seed):
# random.seed(seed)
# os.environ['PYTHONHASHSEED'] = str(seed)
# np.random.seed(seed)
# torch.manual_seed(seed)
def seed_everything(seed):
torch.manual_seed(seed)
if torch.cuda.is_available():
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed) # if you are using multi-GPU.
np.random.seed(seed) # Numpy module.
random.seed(seed) # Python random module.
torch.backends.cudnn.benchmark = False
torch.backends.cudnn.deterministic = True
def main(i):
# configpath =r"D:\codes\codes\stf\mlp\v8\v8_test_config.yaml"
configpath ="/home/hllu/codes/stf/mlp/v8/v8_test_config.yaml"
parser = argparse.ArgumentParser()
parser.add_argument('--cfg', type=str, default=configpath, help="...") # a.yaml中内容在文章开始给出
args = parser.parse_args()
configs = utilme.test_argparse(args)
logpath = configs.logpath
print("logpath is : %s"%logpath)
version = configs.version
# -----------------------------------------------------
# times = time.strftime('%Y%m%d%H%M', time.localtime(time.time()))
timess = time.asctime(time.localtime(time.time()))
print("#####start time: %s######"%timess)
idx = i
checkpoint_path = configs.checkpoint_path
ckp_name = str(version) + "_idx_" + str(idx) + "_.pth"
# ckp_name = str(version) + "_train_v2__idx_" + str(idx) + "_100.pth"
checkpoint_path = os.path.join(checkpoint_path, ckp_name)
# txtfile = configs.txtfile
# checkpoints_name =configs.checkpoints_nameS
# checkpoint_path = os.path.join(configs.logpath, checkpoints_name)
batch_size = configs.batch_size
patchrow = configs.img_size
patchcol = configs.img_size
channels = configs.outchannel
overlap = 0
delta = 0.2
alpha = 0.0
seed = configs.seed
device = 'cuda:1' if torch.cuda.is_available() else 'cpu'
seed_everything(seed)
imglistdir = configs.imglistdir
datalist = utilme.readdatasist(imglistdir)
model = StfNet(configs).to(device)
# print(model)
if os.path.exists(checkpoint_path):
# model.load_state_dict(torch.load(checkpoint_path, map_location=torch.device('cpu')))
# model.load_state_dict(torch.load(checkpoint_path, map_location=torch.device(device)))
model.load_state_dict({k.replace('module.',''):v for k,v in torch.load(checkpoint_path, map_location=torch.device(device)).items()})
print(" Success to loading model dict from %s ....."%checkpoint_path)
else:
print(" Failed to load model dict from %s ....."%checkpoint_path)
return
# print(model)
# if torch.cuda.device_count() > 1:
# # model = torch.nn.DataParallel(model, device_ids=[0, 1, 2])
# model = torch.nn.DataParallel(model, device_ids=[1])
# # model = torch.nn.DataParallel(model)
#-----------------------------------------------------
resultname = "stfmlp_"+datalist[idx][1]
save_path = os.path.join(logpath, resultname)
traindata = MyDataset(configs.trainpath, datalist[idx], patchrow, patchcol,overlap=overlap)
temdata = traindata.__getitem__(0)
#修改这部分代码，此部分代表论文原文中取整幅图像差值的平均值，进行权重的计算。改成对每一个分块分别计算权重。所以改写到下面For循环之内。
v12 = temdata["v12"]
v23 = temdata["v23"]
if (v23 - v12) > delta:
alpha = 1.0
elif (v12 - v23) > delta:
alpha = 0.0
else:
alpha = (1/v12)/(1/v12+1/v23)
# alpha = 1.0
# mybate = v23 / (v12 + v23)
print("delta:%s"%delta)
print("v23-v12:%s"%(v23-v12))
print("v12-v23:%s" % (v12 - v23))
# print(mybate)
print(alpha)
realr = temdata['r']
realc = temdata['c']
row = temdata['paddr']
col = temdata['paddc']
img = torch.zeros((channels,row,col))
train_loader = DataLoader(dataset=traindata, batch_size=batch_size, shuffle=False)
with torch.no_grad():
model.eval()
for i, data in enumerate(train_loader):
idxx = data["idx"].to(device)
c1 = data["c1"].to(device)
c2 = data["c2"].to(device)
c3 = data["c3"].to(device)
f1 = data["f1"].to(device)
f3 = data["f3"].to(device)
c1 = c1.to(torch.float32) / 10000
c2 = c2.to(torch.float32) / 10000
c3 = c3.to(torch.float32) / 10000
f1 = f1.to(torch.float32) / 10000
f3 = f3.to(torch.float32) / 10000
# label = label.to(torch.float32)/10000
x1, x2, f12, f23 = model(c1, c2, c3, f1, f3)
# f2 = 0.5 * (f1 + f12*alpha) + 0.5*(f3 - f23*(1-alpha))
# tem = self.mybate * x4 + (1 - self.mybate) * x3
# f12 = (1 - mybate) * f12
# f23 = mybate * f23
f2 = alpha * (f1 + f12) + (1-alpha) * (f3 - f23)
f2 = torch.squeeze(f2, dim=0)
for i, f2i in enumerate(f2):
rnum = row / patchrow
cnum = col / patchcol
#banchsize>1时使用以下语句
tem = idxx[i]
# tem = idxx
idr = int(tem // cnum)
idc = int(tem % cnum)
idrstart = patchrow * idr
idrend = patchrow * idr + patchrow
idcstart = patchcol * idc
idcend = patchcol * idc + patchcol
if (idrstart - overlap) >= 0:
idrstart -= overlap
idrend += overlap
if (idcstart - overlap) >= 0:
idcstart -= overlap
idcend += overlap
# 无重叠或者重叠区域直接覆盖
img[:, idrstart:idrend, idcstart:idcend] = f2i
out = img[:, 0:realr, 0:realc].numpy()*10000
out = out.astype('int16')
driver = gdal.GetDriverByName("GTiff")
utilme.save_tif2(out, None, None, save_path, driver)
times = time.asctime(time.localtime(time.time()))
print("#####finish time: %s######"%times)
if __name__ == "__main__":
# idinit = 15
# for i in range(idinit):
# print("idx %s"%i)
# main(i)
main(0)