-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathinfer.py
202 lines (165 loc) · 9.52 KB
/
infer.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
import os
import torch
from torch.autograd import Variable
from torchvision import transforms
from utils.config import dataset_root_test
from utils.misc import check_mkdir
from model.MDCNet import MDCNet_fcn,Region_seg
import ttach as tta
import torch.nn.functional as F
import cv2
import numpy as np
torch.manual_seed(2018)
torch.cuda.set_device(0)
ckpt_path = '' # model_pth
anchor_img_path = '' # input_anchor_path
args = {
'snapshot1': 'Region_seg',
'snapshot2': 'MDCNet_seg',
'crf_refine': False,
'save_results': True
}
img_transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
depth_transform = transforms.ToTensor()
target_transform = transforms.ToTensor()
to_pil = transforms.ToPILImage()
to_test = {'PBD_test':dataset_root_test}
transforms = tta.Compose(
[
tta.HorizontalFlip(),
tta.Scale(scales=[0.75,1,1.25], interpolation='bilinear', align_corners=False),
]
)
def Resize(image,H, W):
image = cv2.resize(image, dsize=(W, H), interpolation=cv2.INTER_LINEAR)
return image
def get_coonected_componet_status(img):
img = np.array(img, dtype="uint8")
img = img // 255
retval, labels_cv, stats, centroids = cv2.connectedComponentsWithStats(
img, ltype=cv2.CV_32S
)
return stats
def get_pts_from_xywh(xywh_list):
centroids = []
for item in xywh_list:
x = int(item[0] + item[2] / 2)
y = int(item[1] + item[3] / 2)
centroids.append([x, y])
centroids = np.asarray(centroids, dtype=np.int32).reshape(-1, 2)
centroids = centroids[np.argsort(centroids[:, 1])]
return centroids.tolist()
def cal_maxS_ringS(bubble_img):
bubble_img_np = np.array(bubble_img,dtype = "uint8")
bubble_img_np = bubble_img_np // 255
retval, labels_cv, stats, centroids = cv2.connectedComponentsWithStats(bubble_img_np, ltype=cv2.CV_32S) #计算阈，包括背景
counts_bubble_pixel_fromconnected_regin = [x[-1] for x in stats]
max_index = counts_bubble_pixel_fromconnected_regin.index(max(counts_bubble_pixel_fromconnected_regin[1:]))
x,y, w, h ,s= stats[max_index]
return x,y, w, h ,s
def Normalize(image, mean, std):
image = (image - mean) / std
return image
def main():
net1 = Region_seg().cuda()
net2 = MDCNet_fcn().cuda()
print ('load snapshot \'%s\' for testing' % args['snapshot2'])
net1.load_state_dict(torch.load(os.path.join(ckpt_path, args['snapshot1']+'.pth'),map_location={'cuda:1': 'cuda:1'}))
net2.load_state_dict(torch.load(os.path.join(ckpt_path, args['snapshot2']+'.pth'),map_location={'cuda:1': 'cuda:1'}))
net1.eval()
net2.eval()
with torch.no_grad():
for name, root in to_test.items():
check_mkdir(os.path.join(ckpt_path, '%s_%s' % (name, args['snapshot2'])))
root1 = os.path.join(root, 'img')
img_list = [f for f in os.listdir(root1)]
for idx, img_name in enumerate(img_list):
print ('predicting for %s: %d / %d' % (name, idx + 1, len(img_list)))
rgb_path = os.path.join(root, 'img', img_name)
img = cv2.imread(rgb_path, cv2.IMREAD_COLOR)[:, :, ::-1]
anchor_img = cv2.imread(anchor_img_path, cv2.IMREAD_COLOR)[:, :, ::-1] # anchor
w_,h_,_ = img.shape
w_anchor,h_anchor,_ = anchor_img.shape
img_resize = Resize(img,352,352)
anchor_img_resize = Resize(anchor_img,352,352)
img_var = Variable(img_transform(img_resize).unsqueeze(0), volatile=True).cuda()
anchor_img_var = Variable(img_transform(anchor_img_resize).unsqueeze(0), volatile=True).cuda()
n, c, h, w = img_var.size()
model_output = net1(img_var)
anchor_output = net1(anchor_img_var)
prediction = model_output.sigmoid()
anchor_output = anchor_output.sigmoid()
res = F.upsample(prediction, size=[w_, h_], mode='bilinear', align_corners=False)
res = res.data.cpu().numpy().squeeze()
res = 255 * (res - res.min()) / (res.max() - res.min() + 1e-8)
res[res >= 128] = 255
res[res != 255] = 0
anchor_output = F.upsample(anchor_output, size=[w_anchor, h_anchor], mode='bilinear', align_corners=False)
anchor_output = anchor_output.data.cpu().numpy().squeeze()
anchor_output = 255 * (anchor_output - anchor_output.min()) / (anchor_output.max() - anchor_output.min() + 1e-8)
anchor_output[anchor_output >= 128] = 255
anchor_output[anchor_output != 255] = 0
crop_x, crop_y, crop_w, crop_h, s = cal_maxS_ringS(res)
crop_x_anchor, crop_y_anchor, crop_w_anchor, crop_h_anchor, s_anchor = cal_maxS_ringS(anchor_output)
img_crop = img[crop_y:crop_y + crop_h, crop_x:crop_x + crop_w]
anchor_img_crop = anchor_img[crop_y_anchor:crop_y_anchor + crop_h_anchor, crop_x_anchor:crop_x_anchor + crop_w_anchor]
w_crop, h_crop, _ = img_crop.shape
img_crop_resize = Resize(img_crop, 352, 352)
anchor_img_crop_resize = Resize(anchor_img_crop, 352, 352)
img_crop_var = Variable(img_transform(img_crop_resize).unsqueeze(0), volatile=True).cuda()
anchor_img_crop_var = Variable(img_transform(anchor_img_crop_resize).unsqueeze(0), volatile=True).cuda()
n, c, h, w = img_crop_var.size()
mask_neg = []
mask_pos = []
for transformer in transforms: # custom transforms or e.g. tta.aliases.d4_transform()
rgb_trans = transformer.augment_image(img_crop_var)
rgb_trans_anchor = transformer.augment_image(anchor_img_crop_var)
print(rgb_trans.shape,rgb_trans_anchor.shape)
model_output = net2(rgb_trans,rgb_trans_anchor)
model_output_neg = model_output[:, 0, :, :].unsqueeze(0)
model_output_pos = model_output[:, 1, :, :].unsqueeze(0)
deaug_mask_neg = transformer.deaugment_mask(model_output_neg)
deaug_mask_pos = transformer.deaugment_mask(model_output_pos)
mask_neg.append(deaug_mask_neg)
mask_pos.append(deaug_mask_pos)
output_fpn_neg = torch.mean(torch.stack(mask_neg, dim=0), dim=0)
output_fpn_pos = torch.mean(torch.stack(mask_pos, dim=0), dim=0)
prediction_neg_crop = output_fpn_neg.sigmoid()
prediction_pos_crop = output_fpn_pos.sigmoid()
prediction_neg_crop = F.upsample(prediction_neg_crop, size=[w_crop, h_crop], mode='bilinear', align_corners=False)
prediction_pos_crop = F.upsample(prediction_pos_crop, size=[w_crop, h_crop], mode='bilinear', align_corners=False)
prediction_neg_crop = prediction_neg_crop.data.cpu().numpy().squeeze()
prediction_pos_crop = prediction_pos_crop.data.cpu().numpy().squeeze()
prediction_neg_crop = 255 * prediction_neg_crop
prediction_pos_crop = 255 * prediction_pos_crop
prediction_neg_crop[prediction_neg_crop > 128] = 255
prediction_neg_crop[prediction_neg_crop != 255] = 0
prediction_pos_crop[prediction_pos_crop > 128] = 255
prediction_pos_crop[prediction_pos_crop != 255] = 0
#
prediction_original_neg = np.zeros((img.shape[0], img.shape[1]), np.uint8)
prediction_original_pos = np.zeros((img.shape[0], img.shape[1]), np.uint8)
prediction_original_neg[crop_y:crop_y + crop_h, crop_x:crop_x + crop_w] = prediction_neg_crop
prediction_original_neg[res != 255] = 0
prediction_original_pos[crop_y:crop_y + crop_h, crop_x:crop_x + crop_w] = prediction_pos_crop
prediction_original_pos[res != 255] = 0
prediction_neg_location = prediction_original_neg.copy()
prediction_pos_location = prediction_original_pos.copy()
prediction_neg_location_stats = get_coonected_componet_status(prediction_neg_location)
prediction_pos_location_stats = get_coonected_componet_status(prediction_pos_location)
neg_list = get_pts_from_xywh(prediction_neg_location_stats[1:])
pos_list = get_pts_from_xywh(prediction_pos_location_stats[1:])
if args['save_results']:
check_mkdir(os.path.join(ckpt_path,args['snapshot2']+'epoch',name,'neg_location'))
check_mkdir(os.path.join(ckpt_path,args['snapshot2']+'epoch',name,'pos_location'))
check_mkdir(os.path.join(ckpt_path,args['snapshot2']+'epoch',name,'neg_point_mask'))
check_mkdir(os.path.join(ckpt_path,args['snapshot2']+'epoch',name,'pos_point_mask'))
cv2.imwrite(os.path.join(ckpt_path ,args['snapshot2']+'epoch',name, 'neg_point_mask',img_name[:-4] + '.png'), prediction_original_neg)
cv2.imwrite(os.path.join(ckpt_path ,args['snapshot2']+'epoch',name, 'pos_point_mask',img_name[:-4] + '.png'), prediction_original_pos)
np.save(os.path.join(ckpt_path ,args['snapshot2']+'epoch',name,'neg_location', img_name[:-4] + '.npy'), neg_list)
np.save(os.path.join(ckpt_path ,args['snapshot2']+'epoch',name, 'pos_location',img_name[:-4] + '.npy'), pos_list)
if __name__ == '__main__':
main()