-
Notifications
You must be signed in to change notification settings - Fork 26
/
test.py
131 lines (112 loc) · 5.38 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
import argparse
import numpy as np
from torch.utils.data import Dataset
from torch.utils.data import DataLoader
from models.deeplabv3_version_1.deeplabv3 import DeepLabV3
from torch.autograd import Variable
import torch
import os
import pandas as pd
from PIL import Image
import cv2 as cv
from collections import OrderedDict
import torch.nn as nn
from torchvision import transforms
img_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([.485, .456, .406], [.229, .224, .225])])
def snapshot_forward(model, dataloader, model_list, png):
model.eval()
for (index, (image, pos_list)) in enumerate(dataloader):
image = Variable(image).cuda()
# print(image)
# print(pos_list)
predict_list = 0
for model in model_list:
predict_1 = model(image)
# predict_list = predict_1
predict_2 = model(torch.flip(image, [-1]))
predict_2 = torch.flip(predict_2, [-1])
predict_3 = model(torch.flip(image, [-2]))
predict_3 = torch.flip(predict_3, [-2])
predict_4 = model(torch.flip(image, [-1, -2]))
predict_4 = torch.flip(predict_4, [-1, -2])
predict_list += (predict_1 + predict_2 + predict_3 + predict_4)
predict_list = torch.argmax(predict_list.cpu(), 1).byte().numpy() # n x h x w
batch_size = predict_list.shape[0] # batch大小
for i in range(batch_size):
predict = predict_list[i]
pos = pos_list[i, :]
[topleft_x, topleft_y, buttomright_x, buttomright_y] = pos
if (buttomright_x - topleft_x) == 512 and (buttomright_y - topleft_y) == 512:
png[topleft_y + 128:buttomright_y - 128, topleft_x + 128:buttomright_x - 128] = predict[128:384,128:384]
else:
raise ValueError(
"target_size!=512， Got {},{}".format(buttomright_x - topleft_x, buttomright_y - topleft_y))
h, w = png.shape
png = png[128:h - 128, 128:w - 128] # 去除整体外边界
zeros = (6800, 7200) # 去除补全512整数倍时的右下边界
png = png[:zeros[0], :zeros[1]]
return png
def parse_args():
parser = argparse.ArgumentParser(description="膨胀预测")
parser.add_argument('--test-data-root', type=str, default=r'C:\Users\hekai\Desktop\github-repo\High-Resolution-Remote-Sensing-Semantic-Segmentation-PyTorch\tools\data\image')
parser.add_argument('--test-batch-size', type=int, default=16, metavar='N', help='batch size for testing (default:16)')
parser.add_argument('--num_workers', type=int, default=2)
parser.add_argument("--model-path", type=str, default=r"C:\Users\hekai\Desktop\github-repo\High-Resolution-Remote-Sensing-Semantic-Segmentation-PyTorch\models\pretrained_model\epoch_7_acc_0.84371_kappa_0.77566.pth")
parser.add_argument("--pred-path", type=str, default="")
args = parser.parse_args()
return args
def create_png():
zeros = (6800, 7200)
h, w = zeros[0], zeros[1]
new_h, new_w = (h//512+1)*512, (w//512+1)*512 # 填充下边界和右边界得到滑窗的整数倍
zeros = (new_h+128, new_w+128) # 填充空白边界，考虑到边缘数据
zeros = np.zeros(zeros, np.uint8)
return zeros
class Inference_Dataset(Dataset):
def __init__(self, root_dir, csv_file, transforms):
self.root_dir = root_dir
self.csv_file = pd.read_csv(csv_file, header=None)
self.transforms = transforms
def __len__(self):
return len(self.csv_file)
def __getitem__(self, idx):
filename = self.csv_file.iloc[idx, 0]
# print(filename)
image_path = os.path.join(self.root_dir, filename)
# image = np.asarray(Image.open(image_path)) # mode:RGBA
# image = cv.cvtColor(image, cv.COLOR_RGBA2BGRA) # PIL(RGBA)-->cv2(BGRA)
image = Image.open(image_path).convert('RGB')
# if self.transforms:
# print('transforms')
image = self.transforms(image)
pos_list = self.csv_file.iloc[idx, 1:].values.astype("int") # ---> (topleft_x,topleft_y,buttomright_x,buttomright_y)
return image, pos_list
def reference():
args = parse_args()
dataset = Inference_Dataset(root_dir=args.test_data_root, csv_file=r'C:\Users\hekai\Desktop\github-repo\High-Resolution-Remote-Sensing-Semantic-Segmentation-PyTorch\tools\data\GF2_PMS2_E115.9_N39.0_20170225_L1A0002206085-MSS2.csv', transforms=img_transform)
dataloader = DataLoader(dataset=dataset, batch_size=1, shuffle=False, num_workers=4)
model = DeepLabV3(num_classes=6)
state_dict = torch.load(args.model_path)
new_state_dict = OrderedDict()
for k, v in state_dict.items():
name = k[7:]
new_state_dict[name] = v
model.load_state_dict(new_state_dict)
model = model.cuda()
model = nn.DataParallel(model, device_ids=list(range(torch.cuda.device_count())))
model_list = []
model_list.append(model)
zeros = create_png()
image = snapshot_forward(model, dataloader, model_list, zeros)
label_save_path = "vis_image_" + "_predict.png"
from palette import colorize_mask
overlap=colorize_mask(image)
overlap.save(label_save_path)
img = Image.open(r'E:\data_xiongan\src\GF2_PMS2_E115.9_N39.0_20170225_L1A0002206085-MSS2.jpg').convert('RGBA')
overlap = overlap.convert('RGBA')
image = Image.blend(img, overlap, 0.3)
image.save('blend.png')
if __name__ == '__main__':
reference()