TransCS/eval.py at master · ICSResearch/TransCS · 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
138
139
140
141
142
143
144
import os
import time
import torch
import argparse
import torchvision
import numpy as np
import scipy.io as scio
from skimage.metrics import structural_similarity as SSIM

import utils
import models


parser = argparse.ArgumentParser(description="Args of this repo.")
parser.add_argument("--rate", default=0.1, type=float)
parser.add_argument("--device", default="0")
opt = parser.parse_args()
opt.device = "cuda:" + opt.device


def evaluate():
    print("Start evaluate...")
    config = utils.GetConfig(ratio=opt.rate, device=opt.device)
    net = models.HybridNet(config).to(config.device).eval()

    if os.path.exists(config.model):
        if torch.cuda.is_available():
            trained_model = torch.load(config.model, map_location=config.device)
        else:
            trained_model = torch.load(config.model, map_location="cpu")

        net.load_state_dict(trained_model)
        print("Trained model loaded.")
    else:
        raise FileNotFoundError("Missing trained models.")

    res(config, net, save_img=False)


def res(config, net, save_img):
    tensor2image = torchvision.transforms.ToPILImage()
    save_img = save_img
    batch_size = 1

    net = net.eval()
    file_no = [
        5,
        100,
        100,
    ]

    folder_name = [
        "SET5",
        "BSD100",
        "URBAN100",
    ]

    for idx, item in enumerate(folder_name):
        p_total = 0
        s_total = 0
        mse_total = 0
        path = "{}/".format(config.test_path) + item
        print("*", ("  test dataset: " + path + ", device: " + str(config.device) + "  ").center(120, "="), "*")
        with torch.no_grad():
            for i in range(file_no[idx]):
                name = "{}/({}).mat".format(path, i + 1)
                x = scio.loadmat(name)['temp3']
                x = torch.from_numpy(np.array(x)).to(config.device)
                x = x.float()
                ori_x = x

                h = x.size()[0]
                h_lack = 0
                w = x.size()[1]
                w_lack = 0

                if h % config.block_size != 0:
                    h_lack = config.block_size - h % config.block_size
                    temp_h = torch.zeros(h_lack, w).to(config.device)
                    h = h + h_lack
                    x = torch.cat((x, temp_h), 0)

                if w % config.block_size != 0:
                    w_lack = config.block_size - w % config.block_size
                    temp_w = torch.zeros(h, w_lack).to(config.device)
                    w = w + w_lack
                    x = torch.cat((x, temp_w), 1)

                x = torch.unsqueeze(x, 0)
                x = torch.unsqueeze(x, 0)

                idx_h = range(0, h, config.block_size)
                idx_w = range(0, w, config.block_size)
                num_patches = h * w // (config.block_size ** 2)

                temp = torch.zeros(num_patches, batch_size, config.channel, config.block_size, config.block_size)
                count = 0
                for a in idx_h:
                    for b in idx_w:

                        ori = x[:, :, a:a + config.block_size, b:b + config.block_size].to(config.device)
                        output = net(ori)
                        temp[count, :, :, :, :, ] = output
                        count = count + 1

                y = torch.zeros(batch_size, config.channel, h, w)
                count = 0
                for a in idx_h:
                    for b in idx_w:
                        y[:, :, a:a + config.block_size, b:b + config.block_size] = temp[count, :, :, :, :]
                        count = count + 1

                recon_x = y[:, :, 0:h - h_lack, 0:w - w_lack]

                recon_x = torch.squeeze(recon_x).to("cpu")
                ori_x = ori_x.to("cpu")

                mse = np.mean(np.square(recon_x.numpy() - ori_x.numpy()))
                mse_total += mse
                p = 10 * np.log10(1 / mse)
                p_total = p_total + p

                ssim = SSIM(recon_x.numpy(), ori_x.numpy(), data_range=1)
                s_total = s_total + ssim

                print("\r=> process {:2} done! PSNR: {:5.2f}, SSIM: {:5.4f}, name: {}"
                      .format(i + 1, p, ssim, name))

                if save_img:
                    img_path = "./results/image/{}/".format(int(config.ratio * 100))
                    if not os.path.isdir("./results/image/"):
                        os.mkdir("./results/image/")
                    if not os.path.isdir(img_path):
                        os.mkdir(img_path)
                        print("\rMkdir {}".format(img_path))
                    recon_x = tensor2image(recon_x)
                    recon_x.save(img_path + "({})_{}_{}.png".format(i, p, ssim))

            print("=> All the {:2} images done!, your AVG PSNR: {:5.2f}, AVG SSIM: {:5.4f}"
                  .format(file_no[idx], p_total / file_no[idx], s_total / file_no[idx]))


if __name__ == "__main__":
    evaluate()