/
test.py
149 lines (117 loc) · 5.32 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
import time
import argparse
from osgeo import gdal, gdal_array
import torch
import torchvision.transforms as transforms
import random
from torch.multiprocessing import Process
from pathlib import Path
from torch import nn
import torch.nn.functional as F
import numpy as np
import torch.optim as optim
from torchmetrics import Accuracy, Precision, Recall, F1Score
from model.model import SSNet
import torch.utils.data
from data.data import *
import data.DFC2019Loader as DA
def get_transform(data):
normal_mean_var = {'mean': [0.485, 0.456, 0.406],
'std': [0.229, 0.224, 0.225]}
data = torch.from_numpy(data).float()
transform = transforms.Compose([transforms.Normalize(**normal_mean_var)])
return transform(data).float()
def test(left, right, model, device):
start = time.time()
model.train()
left_tensor = torch.tensor(left, device=device).float()
right_tensor = torch.tensor(right, device=device).float()
with torch.no_grad():
disp1, disp2, disp3, cls1 = model(left_tensor, right_tensor)
print("all", time.time()-start, disp3.shape, cls1.shape)
pred_disp = disp3.detach().cpu().numpy()
cls1 = torch.softmax(cls1[0], dim=0)
pred_label = torch.max(cls1, dim=0)[1]
print(pred_label.shape)
pred_label = pred_label.data.detach().cpu().numpy()
return pred_disp, pred_label
def transform(img):
img = img.GetRasterBand(1)
nodata_value = img.GetNoDataValue()
img = img.ReadAsArray(200, 200, 1024, 1024)
valid = img != nodata_value
avg = np.mean(img[valid])
std = np.std(img[valid])
maxEle = avg + 3 * std
minEle = avg - 3 * std
img1 = img[:]
img1 = np.clip((img1.astype(np.float32) - minEle) /
(maxEle - minEle + 0.00001), 0, 1) * 255.
img = img.astype(np.uint8)
if nodata_value is not None:
img1[img == nodata_value] = 255
img = np.expand_dims(img, axis=0)
img = np.concatenate([img, img, img], axis=0)
print(img.shape)
return img
def main():
start_time = time.time()
parser = argparse.ArgumentParser(description='S3Net')
parser.add_argument('--DFC2019', default='2019', help='DFC2019')
parser.add_argument('--ImgL', default="./data/dataset/", help='ImgL')
parser.add_argument('--ImgR', default="./data/dataset/", help='ImgR')
parser.add_argument('--epochs', default=120, type=int, help='train epoch')
parser.add_argument('--maxdisp', default=48, help='maxium disparity')
parser.add_argument('--model', default='SSNet', help='select model')
parser.add_argument('--train_num', default=700, help='train number')
parser.add_argument('--classfication', default=6, help='class number')
parser.add_argument('--no_cuda', action='store_true', default=False, help='enables CUDA training')
parser.add_argument('--savepath', default="./ckpt/", help='saveckpt')
parser.add_argument('--output', default="./output/", help='save_output')
parser.add_argument('--ckpt', default="./", help='ckpt')
args = parser.parse_args()
args.cuda = not args.no_cuda and torch.cuda.is_available()
device = torch.device("cuda" if args.cuda else "cpu")
print(device, flush=True)
if args.model == "SSNet":
model = SSNet(args.maxdisp, args.classfication)
else:
raise ValueError("no model")
# model = nn.DataParallel(model)
model.to(device)
state_dict = torch.load(args.ckpt)
model.load_state_dict(state_dict['state_dict'])
print('Number of model parameters: {}'.format(sum([p.data.nelement() for p in model.parameters()])))
# w, h = 1024, 1024
# th, tw = 256, 256
# off_x = random.randint(0, w - tw)
# off_y = random.randint(0, h - th)
imgL = gdal.Open(args.ImgL).ReadAsArray()/255.0
imgR = gdal.Open(args.ImgR).ReadAsArray()/255.0
# imgL = gdal.Open(r"Z:\临时存放与文件传输\hongshan\GF701_004549_E114.5_N30.4_20200828111926_MUX_01_SC0_0004_2009012377.tif").ReadAsArray(2000, 2000, 1024, 1024)[:3]/2048.0
# imgR = gdal.Open(r"Z:\临时存放与文件传输\hongshan\GF701_004549_E114.5_N30.4_20200828111926_MUX_01_SC0_0004_2009012377.tif").ReadAsArray(1998, 2000, 1024, 1024)[:3]/2048.0
# imgL = imgL[[2,0,1]].copy()
# imgR = imgR[[2,0,1]].copy()
# gdal_array.SaveArray(imgL, "ori_left.tif")
# gdal_array.SaveArray(imgR, "ori_right.tif")
# imgL = gdal.Open(r"D:\浏览器\download\data_scene_flow\testing\image_2\000000_10.png").ReadAsArray(0, 0, 350, 350)/255.0
# imgR = gdal.Open(r"D:\浏览器\download\data_scene_flow\testing\image_3\000000_10.png").ReadAsArray(0, 0, 350, 350)/255.0
# imgL = transform(imgL)
# imgR = transform(imgR)
# gdal_array.SaveArray(imgL, "ori_left.tif")
imgL = get_transform(imgL)
imgR = get_transform(imgR)
imgL = imgL.unsqueeze(0)
imgR = imgR.unsqueeze(0)
# from thop import profile
# flops, params = profile(model, inputs =(imgL.cuda(), imgR.cuda()))
# print('FLOPs = ' + str(flops/1000**3) + 'G')
# print('Params = ' + str(params/1000**2) + 'M')
pred_disp, pre_cls = test(imgL,imgR, model, device)
print("time", time.time()-start_time)
if not os.path.exists(args.output):
os.makedirs(args.output)
gdal_array.SaveArray(pred_disp, str(Path(args.output)/"disp.tif"))
gdal_array.SaveArray(pre_cls, str(Path(args.output)/"cls.tif"))
if __name__ == "__main__":
main()