ACLRNet/train.py at main · aipixel/ACLRNet · 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
import torch
import argparse
from utils import *
from model import *
from torch.autograd import Variable
from torch.utils.data import DataLoader
import torch.backends.cudnn as cudnn
from torchvision.transforms import ToTensor
from skimage.metrics import structural_similarity as sk_cpt_ssim
from skimage.metrics import peak_signal_noise_ratio as sk_cpt_psnr

def cal_psnr(img1, img2):
    """
    Calculate psnr of the img1 and the img2.
    :param img1: numpy array
    :param img2: numpy array
    :return: np.float32
    """
    return sk_cpt_psnr(img1, img2)
def cal_ssim(img1, img2):
    """
    Calculate ssim of the img1 and the img2.
    :param img1: numpy array
    :param img2: numpy array
    :return: np.float32
    """
    img1 = img1.transpose((1, 2, 0))
    img2 = img2.transpose((1, 2, 0))
    return sk_cpt_ssim(img1, img2, multichannel=True)


def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument("--scale_factor", type=int, default=4)
    parser.add_argument('--device', type=str, default='cuda:0')
    parser.add_argument('--batch_size', type=int, default=12)
    parser.add_argument('--lr', type=float, default=2e-4, help='initial learning rate')
    parser.add_argument('--gamma', type=float, default=0.5, help='optimizer')
    parser.add_argument('--start_epoch', type=int, default=0, help='start epoch')
    parser.add_argument('--n_epochs', type=int, default=80, help='number of epochs to train')
    parser.add_argument('--n_steps', type=int, default=30, help='number of epochs to update learning rate')
    parser.add_argument('--trainset_dir', type=str, default='data/train')
    parser.add_argument('--testset_dir', type=str, default='data/test/KITTI2012/')
    parser.add_argument('--model_name', type=str, default='ACLRNet')
    parser.add_argument('--load_pretrain', type=bool, default=False)
    parser.add_argument('--model_path', type=str, default='log/ACLRNet.pth.tar', help='path to the pretrain model')
    return parser.parse_args()


def train(train_loader, cfg):
    net = Net(upscale_factor=cfg.scale_factor, in_nc=3, out_nc=3, ng0=64, ng=24, nbc=4, nb=2).to(cfg.device)
    net.train()
    cudnn.benchmark = True

    if cfg.load_pretrain:
        if os.path.isfile(cfg.model_path):
            model = torch.load(cfg.model_path, map_location=cfg.device)
            net.load_state_dict(model['state_dict'])
            cfg.start_epoch = model["epoch"]
        else:
            print("=> no model found at '{}'".format(cfg.load_model))

    criterion_L1 = torch.nn.L1Loss().to(cfg.device)
    criterion_L2 = torch.nn.MSELoss().to(cfg.device)
    optimizer = torch.optim.Adam([paras for paras in net.parameters() if paras.requires_grad == True], lr=cfg.lr)

    scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=cfg.n_steps, gamma=cfg.gamma)

    loss_epoch = []
    loss_list = []

    for idx_epoch in range(cfg.start_epoch, cfg.n_epochs):
        for idx_iter, (HR_left, HR_right, LR_left, LR_right) in enumerate(train_loader):
            HR_left, HR_right, LR_left, LR_right  = Variable(HR_left).to(cfg.device), Variable(HR_right).to(cfg.device),\
                                                    Variable(LR_left).to(cfg.device), Variable(LR_right).to(cfg.device)
            SR_left, SR_right = net(LR_left, LR_right, is_training=1)
            loss = criterion_L2(SR_left, HR_left) + criterion_L2(SR_right, HR_right)
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()
            loss_epoch.append(loss.data.cpu())
        scheduler.step()

        if idx_epoch % 1000 == 0:
            psnr_list = []
            ssim_list = []
            psnr_list1 = []
            ssim_list1 = []
            for idx in range(20):
                LR_left = Image.open(cfg.testset_dir + '/lr_x' + str(cfg.scale_factor) + '/' + str(idx+1).zfill(4) + '_L.png')
                LR_right = Image.open(cfg.testset_dir + '/lr_x' + str(cfg.scale_factor) + '/' + str(idx+1).zfill(4) + '_R.png')
                HR_left = Image.open(cfg.testset_dir + '/GT/' + str(idx + 1).zfill(4) + '_L.png')
                HR_right = Image.open(cfg.testset_dir + '/GT/' + str(idx + 1).zfill(4) + '_R.png')
                LR_left, LR_right = ToTensor()(LR_left), ToTensor()(LR_right)
                HR_left, HR_right = ToTensor()(HR_left), ToTensor()(HR_right)
                LR_left, LR_right = LR_left.unsqueeze(0), LR_right.unsqueeze(0)
                HR_left, HR_right = HR_left.unsqueeze(0), HR_right.unsqueeze(0)
                HR_left, HR_right, LR_left, LR_right = Variable(HR_left).to(cfg.device), Variable(HR_right).to(
                    cfg.device), Variable(LR_left).to(cfg.device), Variable(LR_right).to(cfg.device)
                net.eval()
                with torch.no_grad():
                    SR_left, SR_right = net(LR_left, LR_right)
                    SR_left, SR_right = torch.clamp(SR_left, 0, 1), torch.clamp(SR_right, 0, 1)
                if SR_left.shape != HR_left.shape:
                    SR_left = F.interpolate(SR_left, size=(HR_left.shape[2], HR_left.shape[3]), mode='bicubic')
                    SR_right = F.interpolate(SR_right, size=(HR_right.shape[2], HR_right.shape[3]), mode='bicubic')
                psnr_l = cal_psnr(HR_left[:, :, :, :].data.cpu().numpy(), SR_left[:, :, :, :].data.cpu().numpy())
                ssim_l = cal_ssim(HR_left[0, :, :, :].data.cpu().numpy(), SR_left[0, :, :, :].data.cpu().numpy())
                psnr_r = cal_psnr(HR_right[:, :, :, :].data.cpu().numpy(), SR_right[:, :, :, :].data.cpu().numpy())
                ssim_r = cal_ssim(HR_right[0, :, :, :].data.cpu().numpy(), SR_right[0, :, :, :].data.cpu().numpy())
                psnr_list.append(psnr_l)
                ssim_list.append(ssim_l)
                psnr_list1.append(psnr_r)
                ssim_list1.append(ssim_r)
            print('Avg_L. PSNR: {:.5f} dB, Avg. SSIM: {:.5f}'.format(np.mean(psnr_list), np.mean(ssim_list)))
            print('Avg_R. PSNR: {:.5f} dB, Avg. SSIM: {:.5f}'.format(np.mean(psnr_list1), np.mean(ssim_list1)))
            print('Avg. PSNR: {:.5f} dB, Avg. SSIM: {:.5f}'.format(np.mean(psnr_list + psnr_list1), np.mean(ssim_list + ssim_list1)))

        if idx_epoch % 1000 == 0:
            loss_list.append(float(np.array(loss_epoch).mean()))
            print('Epoch--%4d, loss--%f ' % (idx_epoch + 1, float(np.array(loss_epoch).mean())))
            torch.save({'epoch': idx_epoch + 1, 'state_dict': net.state_dict()},
                       './log/' + cfg.model_name + '_' + str(cfg.scale_factor) + 'xSR_epoch' + str(idx_epoch + 1) + '.pth.tar')
            loss_epoch = []


def main(cfg):
    train_set = TrainSetLoader(cfg)
    train_loader = DataLoader(dataset=train_set, num_workers=6, batch_size=cfg.batch_size, shuffle=True)
    train(train_loader, cfg)

if __name__ == '__main__':
    cfg = parse_args()
    main(cfg)