Enhanced_Zero-DCE/train.py at main · Takaluk/Enhanced_Zero-DCE · GitHub

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
import argparse
import os
import warnings

import torch
import torch.optim

import dataloader
import model
import Myloss

# warnings.filterwarnings("ignore", category=UserWarning)

def weights_init(m):
	classname = m.__class__.__name__
	if classname.find('Conv') != -1:
		m.weight.data.normal_(0.0, 0.02)
	elif classname.find('BatchNorm') != -1:
		m.weight.data.normal_(1.0, 0.02)
		m.bias.data.fill_(0)


def train(config):
	img_folder_path = config.lowlight_images_path
	if img_folder_path[-1] == '/':
		img_folder_path = img_folder_path[:len(img_folder_path)-1]

	if config.device == 'cuda':
		device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
		if torch.cuda.is_available():
			os.environ['CUDA_VISIBLE_DEVICES']='0'
		else:
			print('No CUDA GPU found! Using CPU...')
	else:
		device = torch.device('cpu')

	DCE_net = model.enhance_net_nopool().to(device)

	DCE_net.apply(weights_init)
	if config.load_pretrain == True:
		DCE_net.load_state_dict(torch.load(config.pretrain_dir, map_location= device))
	train_dataset = dataloader.lowlight_loader(img_folder_path)
	if len(train_dataset) < 1:
		print('\nNo .jpg file found. Please check your lowlight_images_path.')
		return
	train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=config.train_batch_size, shuffle=True, num_workers=config.num_workers, pin_memory=True)

	L_color = Myloss.L_color()
	L_spa = Myloss.L_spa(device)
	L_exp = Myloss.L_exp(16,0.6, device)
	L_TV = Myloss.L_TV()
	L_lp = Myloss.LPIPSloss(device)
	# L_cont = 	Myloss.ContrastLoss()

	optimizer = torch.optim.Adam(DCE_net.parameters(), lr=config.lr, weight_decay=config.weight_decay)

	DCE_net.train()

	for epoch in range(config.num_epochs):
		for iteration, img_lowlight in enumerate(train_loader):

			img_lowlight = img_lowlight.to(device)

			enhanced_image_, enhanced_image,A  = DCE_net(img_lowlight)

			Loss_TV = 200*L_TV(A)
			loss_spa = torch.mean(L_spa(enhanced_image, img_lowlight))
			loss_col = 5*torch.mean(L_color(enhanced_image))
			loss_exp = 10*torch.mean(L_exp(enhanced_image))
			loss_lp = 3*torch.mean(L_lp(enhanced_image, img_lowlight))
			#loss_cont = 2*torch.mean(L_cont(enhanced_image, img_lowlight))

			# best_loss
			loss =  Loss_TV + loss_spa + loss_col + loss_exp + loss_lp

			optimizer.zero_grad()
			loss.backward()
			torch.nn.utils.clip_grad_norm_(DCE_net.parameters(),config.grad_clip_norm)
			optimizer.step()

		if ((epoch+1) % config.display_iter) == 0:
			print("Loss at epoch", epoch+1, ":", loss.item())

	# 기존에 저장된 가중치 파일 중 가장 큰 숫자를 찾기
	existing_weights = [file for file in os.listdir(config.snapshots_folder) if file.startswith('weight_') and file.endswith('.pth')]
	existing_numbers = [int(file.split('_')[1].split('.')[0]) for file in existing_weights if int(file.split('_')[1].split('.')[0]) != 888 and int(file.split('_')[1].split('.')[0]) != 999]
	if existing_numbers:
		max_existing_number = max(existing_numbers)
	else:
		max_existing_number = -1

	# 888번과 999번을 무시하고 다음 가중치 파일의 숫자를 찾기
	next_number = max(max_existing_number + 1, 0)
	while next_number in existing_numbers or next_number == 888 or next_number == 999:
		next_number += 1

	# 가중치 저장하기
	torch.save(DCE_net.state_dict(), config.snapshots_folder + "/weight_" + str(next_number) + '.pth')

	print(f'\nTrain finished. weight_{str(next_number)}.pth saved to {config.snapshots_folder}.')


if __name__ == "__main__":
	parser = argparse.ArgumentParser()

	# Input Parameters
	parser.add_argument('--device', type=str, default='cuda')
	parser.add_argument('--lowlight_images_path', type=str, default="sample_data/train_data")
	parser.add_argument('--lr', type=float, default=0.0001)
	parser.add_argument('--weight_decay', type=float, default=0.0001)
	parser.add_argument('--grad_clip_norm', type=float, default=0.1)
	parser.add_argument('--num_epochs', type=int, default=200)
	parser.add_argument('--train_batch_size', type=int, default=8)
	parser.add_argument('--val_batch_size', type=int, default=4)
	parser.add_argument('--num_workers', type=int, default=4)
	parser.add_argument('--display_iter', type=int, default=10)
	parser.add_argument('--snapshots_folder', type=str, default="snapshots/")
	parser.add_argument('--load_pretrain', type=bool, default= False)
	parser.add_argument('--pretrain_dir', type=str, default= "snapshots/weight_999.pth")

	config = parser.parse_args()

	if not os.path.exists(config.snapshots_folder):
		os.mkdir(config.snapshots_folder)

	train(config)