-
Notifications
You must be signed in to change notification settings - Fork 1
/
MC_inpainting.py
60 lines (51 loc) · 1.61 KB
/
MC_inpainting.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
'''
Multichannel data inpainting
'''
import numpy as np
from NGR import NGR
from scipy.io import loadmat
### import config file
from config.ngr_inpainting_mc import *
### set data path
## video
data_name = 'claire' # expected result: PSNR=41.05, SSIM=0.990
gt_path = f'./data/{data_name}_gt.mat'
obs_path = f'./data/{data_name}_sr0.15.mat'
## HSI
# data_name = 'PA' # expected result: PSNR=33.72, SSIM=0.961
# gt_path = f'./data/{data_name}_gt.mat'
# obs_path = f'./data/{data_name}_deadline.mat'
### save path
save_path = './result'
gpu_num = 1
if __name__ == '__main__':
### prepare data
gt = loadmat(gt_path)['gt'].transpose(2, 0, 1)
obs = loadmat(obs_path)['obs'].transpose(2, 0, 1)
mask = np.ones_like(obs); mask[obs == 0] = 0
### init NGR
ngr = NGR(obs.astype(np.float32),
gt.astype(np.float32),
mask=mask,
seed=seed,
params=hyperparameters,
save_path=save_path,
show_image=show_image,
gpu_num=gpu_num,
iterations=epoches,
show_every=show_every,
lr=lr,
weight_decay=weight_decay,
exp_weight=exp_weight,
smoothing=smoothing,
data_name=data_name,
task=task,
input_type=input_type,
need_noise_reg=need_noise_reg,
reg_noise_std=reg_noise_std,
input_depth=input_depth,
metrics=metrics)
### train
ngr.train()
### save result
ngr.save()