-
Notifications
You must be signed in to change notification settings - Fork 4
/
LowLightZeroShot.py
149 lines (109 loc) · 4.69 KB
/
LowLightZeroShot.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 torch
import torch.nn as nn
import torchvision
import imageio
import numpy as np
import model
import argparse
import os
import math
import random
from tqdm import tqdm
import warnings
warnings.filterwarnings("ignore")
torch.manual_seed(1)
torch.cuda.manual_seed_all(1)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
np.random.seed(1)
random.seed(1)
def weights_init(m):
classname = m.__class__.__name__
if classname.find('Conv2d') != -1: #0.02
m.weight.data.normal_(0.0, 0.001)
if classname.find('Linear') != -1: #0.02
m.weight.data.normal_(0.0, 0.001)
parser = argparse.ArgumentParser(description='Lowlight Image Enhancement')
parser.add_argument('--TestFolderPath', type=str, default='data/Lowlight/data/LOLTest', help='Lowlight Image folder name')
parser.add_argument('--SavePath', type=str, default='lowresults/LOLTest', help='SavePath Name')
args = parser.parse_args()
def _np2Tensor(img):
np_transpose = np.ascontiguousarray(img.transpose((2, 0, 1)))
tensor = torch.from_numpy(np_transpose).float()
return torch.unsqueeze(tensor, 0)
def psnr(imgS, imgG):
diff = imgS - imgG
mse = diff.pow(2).mean()
return -10 * math.log10(mse)
class I_TV(nn.Module):
def __init__(self):
super(I_TV,self).__init__()
pass
def forward(self,x):
batch_size, h_x, w_x = x.size()[0], x.size()[2], x.size()[3]
count_h, count_w = (h_x-1) * w_x, h_x * (w_x - 1)
h_tv = torch.pow((x[:,:,1:,:]-x[:,:,:h_x-1,:]),2).sum()
w_tv = torch.pow((x[:,:,:,1:]-x[:,:,:,:w_x-1]),2).sum()
return (h_tv/count_h+w_tv/count_w)/batch_size
def randomSelect(_min, _max, _div):
px = random.randint(_min, _max)
px = float(px)/_div
return px
def _augment(_image):
it = random.randint(0, 7)
if it==1: _image = _image.rot90(1, [2, 3])
if it==2: _image = _image.rot90(2, [2, 3])
if it==3: _image = _image.rot90(3, [2, 3])
if it==4: _image = _image.flip(2).rot90(1, [2, 3])
if it==5: _image = _image.flip(3).rot90(1, [2, 3])
if it==6: _image = _image.flip(2)
if it==7: _image = _image.flip(3)
return _image
_img_TV = I_TV()
itr_no = 10000
def test(args):
InputImages = os.listdir(args.TestFolderPath+'/Input/')
os.makedirs(args.SavePath+'/', exist_ok=True)
for i in range(len(InputImages)):
print("Images Processed: %d/ %d \r" % (i+1, len(InputImages)))
_model = model.Model('lowmodel')
_model.apply(weights_init)
_model.cuda()
optimizer = torch.optim.Adam(_model.parameters(), lr=1e-3, betas=(0.99, 0.999), eps=1e-08, weight_decay=1e-2)
Input = imageio.imread(args.TestFolderPath+'/Input/'+InputImages[i])
Input = _np2Tensor(Input)
Input = (Input/255.).cuda()
H, W = Input.shape[2], Input.shape[3]
H, W = H - H%32, W - W%32
Input = Input[:, :, 0:H, 0:W]
for k in tqdm(range(itr_no), desc="Loading..."):
Inputmage = _augment(Input)
_model.train()
optimizer.zero_grad()
trans_map, atm_map, HazefreeImage = _model(Inputmage)
px = 0.9
_trans_map = px
InputmageX = Inputmage*_trans_map + (1 - _trans_map)*atm_map
trans_mapX, atm_mapX, HazefreeImageX = _model(InputmageX)
otensor = torch.ones(HazefreeImage[:, 0:1, :, :].shape).cuda()
ztensor = torch.zeros(HazefreeImage[:, 0:1, :, :].shape).cuda()
lossMxR = torch.sum(torch.max(HazefreeImage[:, 0:1, :, :], otensor)) + torch.sum(torch.max(HazefreeImageX[:, 0:1, :, :], otensor)) - 2*torch.sum(otensor)
lossMxG = torch.sum(torch.max(HazefreeImage[:, 1:2, :, :], otensor)) + torch.sum(torch.max(HazefreeImageX[:, 1:2, :, :], otensor)) - 2*torch.sum(otensor)
lossMxB = torch.sum(torch.max(HazefreeImage[:, 2:3, :, :], otensor)) + torch.sum(torch.max(HazefreeImageX[:, 2:3, :, :], otensor)) - 2*torch.sum(otensor)
lossMx = lossMxR + lossMxG + 10*lossMxB
lossMnR = -torch.sum(torch.min(HazefreeImage[:, 0:1, :, :], ztensor)) - torch.sum(torch.min(HazefreeImageX[:, 0:1, :, :], ztensor))
lossMnG = -torch.sum(torch.min(HazefreeImage[:, 1:2, :, :], ztensor)) - torch.sum(torch.min(HazefreeImageX[:, 1:2, :, :], ztensor))
lossMnB = -torch.sum(torch.min(HazefreeImage[:, 2:3, :, :], ztensor)) - torch.sum(torch.min(HazefreeImageX[:, 2:3, :, :], ztensor))
lossMn = lossMnR + lossMnG + 10*lossMnB
lossT = torch.sum(torch.abs(trans_mapX - px*trans_map))
lossA = torch.sum((atm_map - atm_mapX)**2)
lossTV = _img_TV(HazefreeImage)
loss = lossT + lossA + 0.001*lossMx + 0.01*lossMn + 0.001*lossTV
loss.backward()
optimizer.step()
_model.eval()
with torch.no_grad():
_trans, _atm, _GT = _model(Input)
_GT = torch.clamp(_GT, 0, 1)
torchvision.utils.save_image(_GT, args.SavePath+'/'+InputImages[i][:-4]+'.png')
test(args)