-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path3-predict_scene.py
203 lines (177 loc) · 8.81 KB
/
3-predict_scene.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
# -*- coding: utf-8 -*-
"""
Created on Wed Sep 15 10:20:08 2021
prediction for each image
1 l2flags mask
2 prediction
3 save for each scene
@author: Administrator
"""
# import os
import numpy as np
import h5py
import glob
import os
# troch
import torch
import torch.nn as nn
# import torchvision
from torch.utils.data import Dataset
from tansformer import tnet
from torch.utils.data import DataLoader
from sklearn.metrics import r2_score
from tools import plot_scatter
from neuralnetwork import Net
from L2_flags import L3_mask
from L2wei_QA import QAscores_6Bands
class myDataset(Dataset):
def __init__(self,x,y)-> None:
super().__init__()
self.x_data = x
self.y_data = y
def __getitem__(self, index):
x = self.x_data[index]
y = self.y_data[index]
x_mean = np.array([0.05109411,0.04811902,0.0440184,0.03761746,0.02891396,0.02683822,\
0.02610199,0.02467448,-0.02693318,0.00343917,0.0355856], 'float32')
x_std = np.array([0.01940052,0.01975982,0.01634109,0.02180259,0.01983091,0.0138185,\
0.01695971,0.01088832,0.6893147,0.7120903,0.7063683],'float32')
y_mean = np.array([0.00637908,0.00584464,0.00540286,0.00324083,0.000782,0.00068278],'float32')
y_std = np.array([0.00330699,0.00254451,0.00229242,0.00322304,0.00148145,0.00131122], 'float32')
x = (x-x_mean)/x_std
y = (y-y_mean)/y_std
return x[..., np.newaxis].astype(np.float32).transpose([1, 0]), y[..., np.newaxis].astype(np.float32).transpose([1, 0])
def __len__(self):
return self.x_data.shape[0]
def evaluate(model,val_loader,device='cuda'):
model.eval() # Turn on the evaluation mode
total_label = np.zeros([1,6])
total_predict = np.zeros([1,6])
i = 0
with torch.no_grad():
for x,y in val_loader:
x = x.to(device)
y = y.to(device)
model.to(device)
predict = model(x)
# if i %50 == 0 and i != 0:
# predict = predict.cpu()
# print(str(i+1),'/',str(int(len(val_loader.dataset)/batch_size)))
total_label = np.concatenate((total_label,y.cpu().detach().numpy().squeeze()), axis=0)
total_predict = np.concatenate((total_predict,predict.cpu().detach().numpy().squeeze()),axis=0)
i += 1
total_label = np.delete(total_label,0,0)
total_predict = np.delete(total_predict,0,0)
y_mean = np.array([0.00637908,0.00584464,0.00540286,0.00324083,0.000782,0.00068278],'float32')
y_std = np.array([0.00330699,0.00254451,0.00229242,0.00322304,0.00148145,0.00131122], 'float32')
total_label = total_label*y_std+y_mean
total_predict = total_predict*y_std+y_mean
return total_label, total_predict
if __name__ == "__main__":
n = 0
path = r'D:\2-cloudremove\2-single_scene\G2019276'
files = glob.glob(path+'\*.L2_LAC_OC')
model = torch.load('best_model.pt')
batch_size = 10000
for file in files:
if os.path.exists(file[:-13]+'_Rrs.h5'):
print(file)
else:
'''image read'''
n += 1
print("---第",n,"景---")
data = h5py.File(file,"r")
longitude = data["/navigation_data/longitude"]
[h,w] = longitude.shape
latitude = data["/navigation_data/latitude"]
geod = ['Rrs_412','Rrs_443','Rrs_490','Rrs_555','Rrs_660','Rrs_680',\
'rhos_412','rhos_443','rhos_490','rhos_555','rhos_660','rhos_680',\
'rhos_745','rhos_865',"sena","senz","sola","solz"]
value = np.empty((h*w, len(geod)),'float32')
#2.维度转换（二维转一维），scale、add修正
for i in range(len(geod)):
dataset_band = data["/geophysical_data/" + geod[i]]
value_band = dataset_band[:, :] * 1.
# value_band[value_band == -32767.] = np.nan
value[:,i] = value_band.flatten()
try:
gain = dataset_band.attrs["scale_factor"][0]
offset = dataset_band.attrs["add_offset"][0]
except:
gain = 1
offset = 0
value[:,i] = value[:,i]*gain + offset
'''l2_flags mask'''
flags = [0,1,3,4,5,6,10,12,14,15,16,19,20,21,22,24,25]
flag_land = [1]
# l2flags = np.array(value[:,-1], dtype='int32').transpose()
l2flags = data["/geophysical_data/l2_flags"]
value_masked = L3_mask(flags, l2flags) #good pixel=1
idx_land = L3_mask(flag_land, l2flags) #extract land pixels
delete_thick_cloud_idx = np.argwhere(value[:,-5]<0.06) #delete Rrc865>0.06
val_index = np.argwhere(value_masked.flatten()==1)
val_index = list(set(val_index.squeeze()).intersection(set(delete_thick_cloud_idx.squeeze())))
data_masked = value[val_index,:]
'''cal RAA'''
train_data = data_masked[:,6:] #0-7：rrs, 8-21:rrc,22-26:geometry
train_label = data_masked[:,:6]
train_dataset = train_data[:,:-1]
train_dataset[:,-3] = np.cos(train_data[:,-3])
train_dataset[:,-2] = np.cos(train_data[:,-1])
train_dataset[:,-1] = abs(train_data[:,-4]-train_data[:,-2])
idx1 = np.argwhere(train_dataset[:,10]>180)
column_ra = train_dataset[:,-1]
column_ra[idx1.squeeze()] = 360-column_ra[idx1.squeeze()]
train_dataset[:,-1] = column_ra
train_dataset[:,-1]=np.cos(train_dataset[:,-1])
'''prediction'''
target_dt = myDataset(train_dataset,train_label)
val_loader = DataLoader(target_dt, shuffle=False, batch_size=batch_size,
num_workers=0, drop_last=False, pin_memory=True)
label, predict = evaluate(model, val_loader)
'''QA'''
total_score_sd = []
total_score_dl = []
length_td = len(label)
num_batch = round(length_td/10)
test_lambda = np.array([412,443,490,555,660,680])
for batch in range(10):
if batch<9:
test_Rrs_sd = label[batch*num_batch:num_batch*(batch+1),:6]
test_Rrs_dl = predict[batch*num_batch:num_batch*(batch+1),:6]
maxCos_sd, cos_sd, clusterID_sd, totScore_sd = QAscores_6Bands(test_Rrs_sd, test_lambda)
maxCos_dl, cos_dl, clusterID_dl, totScore_dl = QAscores_6Bands(test_Rrs_dl, test_lambda)
total_score_sd = np.concatenate((total_score_sd,totScore_sd))
total_score_dl = np.concatenate((total_score_dl,totScore_dl))
else:
test_Rrs_sd = label[batch*num_batch:,:6]
test_Rrs_dl = predict[batch*num_batch:,:6]
maxCos_sd, cos_sd, clusterID_sd, totScore_sd = QAscores_6Bands(test_Rrs_sd, test_lambda)
maxCos_dl, cos_dl, clusterID_dl, totScore_dl = QAscores_6Bands(test_Rrs_dl, test_lambda)
total_score_sd = np.concatenate((total_score_sd,totScore_sd))
total_score_dl = np.concatenate((total_score_dl,totScore_dl))
'''back to 2-D image'''
image_sd = np.zeros((h,w,6),'float32')
image_dl = np.zeros((h,w,6),'float32')
image_score_sd = np.zeros((h,w),'float32')
image_score_dl = np.zeros((h,w),'float32')
temp_sd = np.zeros((h*w,1),'float32')
temp_dl = np.zeros((h*w,1),'float32')
temp_score = np.zeros((h*w,1),'float32')
'''save'''
f = h5py.File(file[:-10]+'_Rrs.h5','w')
for j in range(6):
temp_sd[val_index,0] = label[:,j]
temp_dl[val_index,0] = predict[:,j]
image_sd[:,:,j] = temp_sd.reshape(h,w)
image_dl[:,:,j] = temp_dl.reshape(h,w)
f.create_dataset('seadas_'+str(j+1),data=image_sd[:,:,j])
f.create_dataset('ACDL_'+str(j+1),data=image_dl[:,:,j])
temp_score[val_index,0] = total_score_sd
image_score_sd = temp_score.reshape(h,w)
f.create_dataset('score_sd',data=image_score_sd)
temp_score[val_index,0] = total_score_dl
image_score_dl = temp_score.reshape(h,w)
f.create_dataset('score_dl',data=image_score_dl)
f.create_dataset('land',data=idx_land)
f.close()