-
Notifications
You must be signed in to change notification settings - Fork 0
/
infer.py
135 lines (104 loc) · 4.86 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
import numpy as np
import os
import torch
from PIL import Image
from torchvision import transforms
from skimage import io
from config import ViSha_test_root
from misc import check_mkdir
from networks.VGD_reflection import VGD_Network
from dataset.VSshadow_ours import listdirs_only
import argparse
from tqdm import tqdm
# os.environ["CUDA_VISIBLE_DEVICES"] = "0"
config = {
'scale': 416,
'test_adjacent': 1,
'input_folder': 'JPEGImages',
'label_folder': 'SegmentationClassPNG'
}
img_transform = transforms.Compose([
transforms.Resize((config['scale'], config['scale'])),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
target_transform = transforms.ToTensor()
root = ViSha_test_root[0]
print('root: ', root)
to_pil = transforms.ToPILImage()
import pdb
parser = argparse.ArgumentParser()
parser.add_argument("-pred", "--prediction", type=str, default=None) #results/
parser.add_argument("-exp", "--exp", type=str, default="VMD_ours")
args = parser.parse_args()
def save_reflection(image_name, pred, d_dir, size):
predict = pred
predict = predict.squeeze()
predict_np = predict.cpu().data.numpy()
predict_np = np.clip(predict_np, 0, 1)
predict_np = np.transpose(predict_np, (1, 2, 0)) * 255.
im = Image.fromarray((predict_np.astype(np.uint8))).convert('RGB')
imo = im.resize(size, resample=Image.BILINEAR)
imo.save(os.path.join(d_dir, image_name + '.png'))
def main():
net = VGD_Network().cuda()
print(args.prediction)
print(args.exp)
# checkpoint = os.path.join(args.exp, 'best.pth')
checkpoint = args.exp
print(checkpoint)
check_point = torch.load(checkpoint)
msg = net.load_state_dict(check_point['model'], strict=False)
print(msg)
import time
all_time = 0
index = 0
net.eval()
with torch.no_grad():
video_list = listdirs_only(os.path.join(root))
# print(video_list)
video_list = sorted(video_list)
for video in tqdm(video_list):
# all images
img_list = [os.path.splitext(f)[0] for f in os.listdir(os.path.join(root, video, config['input_folder']))
if f.endswith('.jpg')]
# need evaluation images
img_eval_list = [os.path.splitext(f)[0] for f in os.listdir(os.path.join(root, video, config['label_folder']))
if f.endswith('.png')]
img_eval_list = sortImg(img_eval_list)
for exemplar_idx, exemplar_name in enumerate(img_eval_list):
query_idx_list = getAdjacentIndex(exemplar_idx, 0, len(img_list), config['test_adjacent'])
for query_idx in query_idx_list:
index += 1
exemplar = Image.open(os.path.join(root, video, config['input_folder'], exemplar_name + '.jpg')).convert('RGB')
w, h = exemplar.size
query = Image.open(os.path.join(root, video, config['input_folder'], img_list[query_idx] + '.jpg')).convert('RGB')
exemplar_tensor = img_transform(exemplar).unsqueeze(0).cuda()
query_tensor = img_transform(query).unsqueeze(0).cuda()
start = time.time()
exemplar_final, exemplar_ref, exemplar_pre = net(exemplar_tensor, query_tensor, query_tensor)
all_time += time.time() - start
# exemplar_final = exemplar_pre #### NOTE
res = (exemplar_final.data > 0).to(torch.float32).squeeze(0)
# res = torch.sigmoid(exemplar_final.squeeze())
prediction = np.array(
transforms.Resize((h, w))(to_pil(res.cpu())))
check_mkdir(os.path.join(args.prediction, 'pred', video))
save_name = f"{exemplar_name}.png"
Image.fromarray(prediction).save(os.path.join(args.prediction, 'pred', video, save_name))
# print(os.path.join(args.prediction, 'reflection', video), '---')
# # ## save reflection
# check_mkdir(os.path.join(args.prediction, 'reflection', video))
# save_reflection(exemplar_name, exemplar_ref, os.path.join(args.prediction, 'reflection', video), (w, h))
def sortImg(img_list):
img_int_list = [int(f) for f in img_list]
sort_index = [i for i, v in sorted(enumerate(img_int_list), key=lambda x: x[1])] # sort img to 001,002,003...
return [img_list[i] for i in sort_index]
def getAdjacentIndex(current_index, start_index, video_length, adjacent_length):
if current_index + adjacent_length < start_index + video_length:
query_index_list = [current_index+i+1 for i in range(adjacent_length)]
else:
query_index_list = [current_index-i-1 for i in range(adjacent_length)]
return query_index_list
if __name__ == '__main__':
main()