-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtools.py
325 lines (291 loc) · 12.7 KB
/
tools.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
# -*- coding: utf-8 -*-
"""
Created on Thu May 20 15:56:29 2021
1. extract sub dataset for test
2. train_datda_analysis
3. earlystopping
4. plot_scatter
@author: Administrator
"""
import h5py
import numpy as np
import glob
import torch
from matplotlib import pyplot as plt
import torch
from sklearn.metrics import r2_score
from scipy.stats import gaussian_kde
def test_data(path):
rate = 0.7
data = h5py.File(path)
x = np.array(data['train'])
x[:,-3:] = np.cos(np.radians(x[:,-3:]))
y = np.array(data['label'])
z = np.array(data['cloud_Rrs'])
# x_mean = np.mean(x,axis=0)
# x_std = np.std(x,axis=0)
# y_mean = np.mean(y,axis=0)
# y_std = np.std(y,axis=0)
# print(x_mean,'\n',x_std,'\n',y_mean,'\n',y_std)
length = len(x)
induces = torch.randperm(length)
xx = x[induces[:int(length*rate)],:]
yy = y[induces[:int(length*rate)],:]
zz = z[induces[:int(length*rate)],:]
xxx = x[induces[int(length*rate):],:]
yyy = y[induces[int(length*rate):],:]
zzz = z[induces[int(length*rate):],:]
print('train mean:',np.mean(x,axis=0))
print('train std:',np.std(x,axis=0))
print('label mean:',np.mean(y,axis=0))
print('label std:',np.std(y,axis=0))
f = h5py.File('train_ds.h5','w')
f.create_dataset('train',data=xx)
f.create_dataset('label',data=yy)
f.create_dataset('cloud_Rrs',data=zz)
f.close()
f1 = h5py.File('test_ds.h5','w')
f1.create_dataset('train',data=xxx)
f1.create_dataset('label',data=yyy)
f1.create_dataset('cloud_Rrs',data=zzz)
f1.close()
def OCI2(band1,band2,band3):
'''
MODIS
band1:443
band2:555
band3:667
'''
ci = band2-0.5*(band1+band3)
ci[np.argwhere(ci>0.004)] = 0.004 #ci指数小于-0.0005的可用于计算chl, 大概范围【-0.008，0.004】
# chl = 10**(-0.4909+191.6590*ci) #OCI 1
chl = 10**(-0.4287+230.47*ci) #OCI 2
return chl.squeeze()
def extract_chl_OCI2(path):
'''根据叶绿素浓度提取数据集'''
thre = 200000
data = h5py.File(path)
x = np.array(data['train'])
y = np.array(data['label'])
chl = OCI2(y[:,1],y[:,5],y[:,6])
levels = [0.0,0.05,0.1,0.15,0.2,0.25,0.3,0.35]
train = np.zeros([1,17])
label = np.zeros([1,8])
for i,level in enumerate(levels):
print(str(i+1),'/',str(len(levels)))
up = level+0.05
low = level
idx = np.array(list(set(np.argwhere(chl>low).flatten()).intersection(set(np.argwhere(chl<up).flatten()))))
if len(idx)>thre:
#随机取thre个数据
rand_num = torch.randperm(len(idx)).numpy()
train = np.concatenate((train,x[idx[rand_num[:thre]],:]),axis=0)
label = np.concatenate((label,y[idx[rand_num[:thre]],:]),axis=0)
else:
print('The number is not enough at:',level)
train = np.delete(train,0,0)
label = np.delete(label,0,0)
f = h5py.File('train_data_OCI2.h5','w')
f.create_dataset('train',data=train)
f.create_dataset('label',data=label)
f.close()
def extract_turbid(path):
'''根据667nm提取数据集'''
thre = 200000
data = h5py.File(path)
x = np.array(data['train'])
y = np.array(data['label'])
turbid = y[:,-2]
levels = [0.0000,0.0001,0.0002,0.0003,0.0004,0.0005,0.0006,0.0007]
train = np.zeros([1,17])
label = np.zeros([1,8])
for i,level in enumerate(levels):
print(str(i+1),'/',str(len(levels)))
up = level+0.0001
low = level
idx = np.array(list(set(np.argwhere(turbid>low).flatten()).intersection(set(np.argwhere(turbid<up).flatten()))))
if len(idx)>thre:
#随机取thre个数据
rand_num = torch.randperm(len(idx)).numpy()
train = np.concatenate((train,x[idx[rand_num[:thre]],:]),axis=0)
label = np.concatenate((label,y[idx[rand_num[:thre]],:]),axis=0)
else:
train = np.concatenate((train,x[idx,:]),axis=0)
label = np.concatenate((label,y[idx,:]),axis=0)
print('The number is not enough at:',level,len(idx))
train = np.delete(train,0,0)
label = np.delete(label,0,0)
f = h5py.File('train_data_667nm.h5','w')
f.create_dataset('train',data=train)
f.create_dataset('label',data=label)
f.close()
def train_data_analysis(path):
files = glob.glob(path+'dataset_*.h5')
i = 0
for file in files:
i+=1
print('part:'+str(i))
data = h5py.File(file)
x = np.array(data['train'])
y = np.array(data['label'])
if i == 1:
#metrics calculating
mean_rrc = np.mean(x,axis=0)[np.newaxis,:] #including geometries
std_rrc = np.std(x,axis=0)[np.newaxis,:]
min_rrc = np.amin(x,axis=0)[np.newaxis,:]
max_rrc = np.amax(x,axis=0)[np.newaxis,:]
mean_rrs = np.mean(y,axis=0)[np.newaxis,:] #including geometries
std_rrs = np.std(y,axis=0)[np.newaxis,:]
min_rrs = np.amin(y,axis=0)[np.newaxis,:]
max_rrs = np.amax(y,axis=0)[np.newaxis,:]
watertype = x[:,-1]
types = np.unique(watertype)
types = types[:,np.newaxis]
for j,k in enumerate(types[:,0]):
types[j,1] = np.sum(watertype==k)
print(types)
else:
mean_rrc1 = np.mean(x,axis=0)[np.newaxis,:] #including geometries
std_rrc1 = np.std(x,axis=0)[np.newaxis,:]
min_rrc1 = np.amin(x,axis=0)[np.newaxis,:]
max_rrc1 = np.amax(x,axis=0) [np.newaxis,:]
mean_rrs1 = np.mean(y,axis=0)[np.newaxis,:] #including geometries
std_rrs1 = np.std(y,axis=0)[np.newaxis,:]
min_rrs1 = np.amin(y,axis=0)[np.newaxis,:]
max_rrs1 = np.amax(y,axis=0)[np.newaxis,:]
mean_rrc = np.concatenate((mean_rrc[:],mean_rrc1),axis=0)
std_rrc = np.concatenate((std_rrc,std_rrc1),axis=0)
min_rrc = np.concatenate((min_rrc,min_rrc1),axis=0)
max_rrc = np.concatenate((max_rrc,max_rrc1),axis=0)
mean_rrs = np.concatenate((mean_rrs,max_rrs1),axis=0)
std_rrs = np.concatenate((std_rrs,std_rrs1),axis=0)
min_rrs = np.concatenate((min_rrs,min_rrs1),axis=0)
max_rrs = np.concatenate((max_rrs,max_rrs1),axis=0)
mean_rrc = np.mean(mean_rrc,axis=0) #including geometries
std_rrc = np.mean(std_rrc,axis=0)
min_rrc = np.amin(min_rrc,axis=0)
max_rrc = np.amax(max_rrc,axis=0)
mean_rrs = np.mean(mean_rrs,axis=0) #including geometries
std_rrs = np.mean(std_rrs,axis=0)
min_rrs = np.amin(min_rrs,axis=0)
max_rrs = np.amax(max_rrs,axis=0)
return mean_rrc,std_rrc,min_rrc,max_rrc,mean_rrs,std_rrs,min_rrs,max_rrs
class EarlyStopping:
#https://blog.csdn.net/qq_37430422/article/details/103638681
"""Early stops the training if validation loss doesn't improve after a given patience."""
def __init__(self, patience=7,verbose=False, delta=0):
"""
Args:
patience (int): How long to wait after last time validation loss improved.
Default: 7
verbose (bool): If True, prints a message for each validation loss improvement.
Default: False
delta (float): Minimum change in the monitored quantity to qualify as an improvement.
Default: 0
"""
self.patience = patience
self.verbose = verbose
self.counter = 0
self.best_score = None
self.early_stop = False
self.val_loss_min = np.Inf
self.delta = delta
def __call__(self, val_loss, model):
score = -val_loss
if self.best_score is None:
self.best_score = score
self.save_checkpoint(val_loss, model)
elif score < self.best_score + self.delta:
self.counter += 1
print(f'EarlyStopping counter: {self.counter} out of {self.patience}')
if self.counter >= self.patience:
self.early_stop = True
else:
self.best_score = score
self.save_checkpoint(val_loss, model)
self.counter = 0
def save_checkpoint(self, val_loss, model):
'''Saves model when validation loss decrease.'''
if self.verbose:
print(f'Validation loss decreased ({self.val_loss_min:.6f} --> {val_loss:.6f}). Saving model ...')
torch.save(model, 'best_model.pt') # 这里会存储迄今最优模型的参数
self.val_loss_min = val_loss
def plot_scatter(x1,y1,stats_dl,band):
'''
x: predict
y: label
'''
min_val = np.amin([x1,y1])
max_val = np.amax([x1,y1])
x0 = np.array([min_val,max_val])
y0 = np.array([min_val,max_val])
# ZX = (mat1-np.mean(mat1))/np.std(mat1)
# ZY = (mat2-np.mean(mat2))/np.std(mat2)
# r = np.sum(ZX*ZY)/(len(mat1))
r2_1 = r2_score(x1,y1)
rpd_1 = np.mean((x1-y1)/y1)
bias_1 = np.mean(y1-x1)
#density scatter
xy1 = np.vstack([x1, y1])
z = gaussian_kde(xy1)(xy1)
idx = z.argsort()
x1, y1, z1 = x1[idx], y1[idx], z[idx]
#创建绘图
fig = plt.figure(num=1, figsize=(3.5, 3.0)) #figsize单位为英寸
ax = plt.subplot(111)
# 设置字体
# plt.rcParams['axes.unicode_minus'] = False #使用上标小标小一字号
# plt.rcParams['font.sans-serif']=['Times New Roman'] #设置全局字体，‘SimHei’黑体可现实中文
font1 = {
'color': 'black',
'weight': 'normal', #wight为字体的粗细，可选 ‘normal\bold\light’等
'size': 10
}
font2 = {
'color': 'black',
'weight': 'normal',
'size': 13
}
font3 = {
'color': 'darkorange',
'weight': 'normal', #wight为字体的粗细，可选 ‘normal\bold\light’等
'size': 5
}
# plt.rc('font', family='Times New Roman', size=7)
#设置x,y轴的风格
# ax.tick_params(axis='x', which='major', bottom=True, top=False, labelbottom=True,
# direction='inout', labelsize=10, width=1, length=10, colors='black')
# ax.tick_params(axis='x', which='minor', bottom=True, top=False, labelbottom=True,
# direction='in', labelsize=10, width=1, length=5,colors='black')
# ax.tick_params(axis='y', which='major', left=True, right=False,labelbottom=True,
# direction='inout', labelsize=10, width=1, length=10,colors='black')
# ax.tick_params(axis='y', which='minor', left=True, right=False,labelbottom=True,
# direction='in', labelsize=10, width=1, length=5,colors='black')
# f1 = ax.plot(x, y, marker='o', markersize=1.2, color='blue', linewidth=0.0, linestyle='--')
plt.xlim(min_val,max_val)
plt.ylim(min_val,max_val)
#绘图
f0 = ax.plot(x0,y0, color='black',linewidth=1)
f1 = ax.scatter(x1,y1, marker='o', s=3, c=z1,cmap='rainbow',linewidth=0.0) #x轴为insitu，y轴predict
# plt.text(0.05,0.9, 'N=%d'%(len(x1)),fontdict=font1,transform=ax.transAxes)
plt.text(0.05,0.9,'MAPD:%6.2f'%(stats_dl[0]*100)+'%',fontdict=font1,transform=ax.transAxes)
plt.text(0.05,0.8, 'R:%6.2f'%(stats_dl[1]),fontdict=font1,transform=ax.transAxes)
plt.text(0.05,0.7, 'bias:%8.4f sr$^{-1}$'%(stats_dl[2]),fontdict=font1,transform=ax.transAxes)
# plt.title(name+band)
# ax.tick_params(axis='y', direction='in', length=3, width=1, colors='black', labelrotation=90)
#设置坐标名
ax.set_ylabel(r'Retrieved R$\mathregular{_{rs}}$'+r' (sr$\mathregular{^{-1}}$)', fontdict=font2)
ax.set_xlabel(r'R$\mathregular{_{rs,true}}$'+r' (sr$\mathregular{^{-1}}$)', fontdict=font2)
# plt.minorticks_on() #开启小坐标
# ax.xaxis.set_label_coords(0.5, -0.11)
# ax.tick_params(axis='x', direction='in', length=3, width=1, colors='black', labelrotation=0)
# plt.ticklabel_format(axis='both',style='sci') #sci文章的风格
plt.tight_layout(rect=(0,0,1,1))#rect=[left,bottom,right,top] #设置图框与图片边缘的距离
figname = r'./GOCI_'+str(band)+'_500dpi.png'
plt.savefig(figname, dpi=600)
plt.close()
if __name__=='__main__':
path = r'GOCI_ds_0623.h5'
# mean_rrc,std_rrc,min_rrc,max_rrc,mean_rrs,std_rrs,min_rrs,max_rrs = train_data_analysis(path)
# print(mean_rrc,std_rrc,min_rrc,max_rrc,mean_rrs,std_rrs,min_rrs,max_rrs)
test_data(path)