AutoUnmix/test.py at main · AlphaYuan/AutoUnmix · 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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
import argparse
import os
import sys
import time

import numpy as np
import torch.nn as nn
import torch.utils.data
from torch.utils.data import DataLoader
from torchvision import transforms
from torchvision.utils import save_image
from skimage.metrics import peak_signal_noise_ratio

import model
import model_new
import pytorch_msssim
from utils import MyDataSet, show_tif

def parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument("--epoch", type=int, default=0, help="epoch to start training from")
    parser.add_argument("--epochs", type=int, default=200, help="number of epochs of training")
    parser.add_argument("--test_epoch", type=int, default=200, help="number of epochs of training")
    parser.add_argument("--dataset_name", type=str, default="synthetic", help="name of the dataset")
    parser.add_argument("--data_path", type=str, default="data/train/mixed", help="path of the dataset")
    parser.add_argument("--target_path", type=str, default="data/train/target", help="path of the target set")
    parser.add_argument("--batch_size", type=int, default=1, help="size of the batches")
    parser.add_argument("--lr", type=float, default=0.0002, help="adam: learning rate")
    parser.add_argument("--b1", type=float, default=0.5, help="adam: decay of first order momentum of gradient")
    parser.add_argument("--b2", type=float, default=0.999, help="adam: decay of first order momentum of gradient")
    parser.add_argument("--decay_epoch", type=int, default=100, help="epoch from which to start lr decay")
    parser.add_argument("--n_cpu", type=int, default=8, help="number of cpu threads to use during batch generation")
    # parser.add_argument("--img_height", type=int, default=512, help="size of image height")
    # parser.add_argument("--img_width", type=int, default=512, help="size of image width")
    parser.add_argument("--re_size", type=int, default=512, help="size of image")
    parser.add_argument("--crop_size", type=int, default=256, help="size of cropped image")
    parser.add_argument("--channels", type=int, default=4, help="number of image channels")
    # parser.add_argument("--sample_interval", type=int, default=100, help="interval between saving generator outputs")
    parser.add_argument("--n_residual_blocks", type=int, default=9, help="number of residual blocks in generator")
    parser.add_argument("--loss", type=str, default="MSELoss", help="loss function")
    parser.add_argument('--local_rank', default=0, type=int, help='node rank for distributed training')
    parser.add_argument("--model", default="ResNet", type=str, help="model for training")
    parser.add_argument("--version", default=0, type=int, help="version")
    parser.add_argument("--augmentation", action='store_false', help="image augmentation")
    opt = parser.parse_args()
    print(opt)
    return opt

opt = parse_args()
torch.cuda.set_device(0)

os.makedirs('images/%s/version_%s/test' % (opt.dataset_name, opt.version), exist_ok=True)

cuda = torch.cuda.is_available()
input_shape = (opt.channels, opt.re_size, opt.re_size)

if cuda:
    try:
        from apex import amp
    except:
        print("Apex is not installed. Use PyTorch DDP")

transform = transforms.Compose([
    # transforms.Resize(opt.re_size),
    transforms.CenterCrop(512),
    transforms.ToTensor(),
])

# train_set = MyDataSet(opt.data_path, opt.target_path, transform, augmentation=True, crop_size=(opt.crop_size, opt.crop_size))
test_set = MyDataSet(opt.data_path.replace('train', 'test'), opt.target_path.replace('train', 'test'), transform, opt.augmentation, (opt.crop_size, opt.crop_size), test_batch4=True)
# train_sampler = torch.utils.data.distributed.DistributedSampler(train_set)
# train_loader = DataLoader(dataset=train_set, batch_size=opt.batch_size, num_workers=opt.n_cpu, pin_memory=True,
#                           sampler=train_sampler)
# train_loader = DataLoader(dataset=train_set, batch_size=opt.batch_size, num_workers=opt.n_cpu, pin_memory=True,)
test_loader = DataLoader(dataset=test_set, batch_size=opt.batch_size, shuffle=False, num_workers=opt.n_cpu, )
print("testset size: {}, testloader size: {}".format(len(test_set), len(test_loader)))

# if opt.model == 'ResNet':
#     enc = model.ResNet(input_shape, opt.n_residual_blocks)
#     dec = model.ResNet(input_shape, opt.n_residual_blocks)
# elif opt.model == 'ResNet_old':
#     enc = model.ResNet_old(input_shape, opt.n_residual_blocks)
#     dec = model.ResNet_old(input_shape, opt.n_residual_blocks)
# elif opt.model == 'CNN':
#     enc = model.Encoder()
#     dec = model.Decoder()
# elif opt.model == 'UNet':
#     enc = model.UNet(input_shape[0], input_shape[0])
#     # enc = model.ResNet(input_shape, opt.n_residual_blocks)
#     dec = model.UNet(input_shape[0], input_shape[0])
#     # dec = model.ResNet(input_shape, opt.n_residual_blocks)
# elif opt.model == 'UwUNet':
#     # enc = model.UwUNet(input_shape[0], input_shape[0])
#     # enc = model.ResNet(input_shape, opt.n_residual_blocks)
#     enc = model.UNet(input_shape[0], input_shape[0])
#     dec = model.UwUNet(input_shape[0], input_shape[0])
# elif opt.model == 'ConvNeXt':
#     enc = model_convnext.ConvNeXt(in_chans=input_shape[0], out_chans=input_shape[0], )
#     dec = model_convnext.ConvNeXt(in_chans=input_shape[0], out_chans=input_shape[0], )
# else:
#     enc = model.GeneratorResNet(input_shape, opt.n_residual_blocks)
#     dec = model.GeneratorResNet(input_shape, opt.n_residual_blocks)
## myNet
enc = model_new.UNet(input_shape[0], input_shape[0])
dec = model_new.unmixingNet(input_shape[0], input_shape[0])

criterion = nn.MSELoss()
criterion1 = nn.L1Loss()
ssim1 = pytorch_msssim.MSSSIM()

if cuda:
    enc = enc.cuda()
    dec = dec.cuda()
    criterion.cuda()
    criterion1.cuda()

epoch = ''
if opt.epochs == opt.test_epoch:
    epoch
elif opt.test_epoch >= 0:
    epoch = '_' + str(opt.test_epoch)
elif opt.test_epoch < 0:
    if opt.test_epoch == -1:
        epoch = '_ft'
    else:
        epoch = '_' + str(abs(opt.test_epoch)) + '_ft'
print("saved_models/%s/enc%s.pth" % (opt.dataset_name, epoch))
pretrained_enc = torch.load("saved_models/%s/version_%s/enc%s.pth" % (opt.dataset_name, opt.version, epoch), map_location='cpu')
print("saved_models/%s/dec%s.pth" % (opt.dataset_name, epoch))
pretrained_dec = torch.load("saved_models/%s/version_%s/dec%s.pth" % (opt.dataset_name, opt.version, epoch), map_location='cpu')

enc.load_state_dict({k.replace('module.', ''): v for k, v in pretrained_enc.items()})
dec.load_state_dict({k.replace('module.', ''): v for k, v in pretrained_dec.items()})

dec = amp.initialize(dec, opt_level='O0')

enc.eval()
dec.eval()

print("start test: ")
time_used = []
avg_loss = [[], [], [], []]
avg_loss_x = [[], [], [], []]
for i, (x, y) in enumerate(test_loader):

    x = x[:,[0,2,3],:,:]
    y = y[:,[0,2,3],:,:]
    x = x[:,:opt.channels,:,:]
    y = y[:,:opt.channels,:,:]
    # print(x.shape, y.shape)
    st = time.time()
    if cuda:
        x = x.cuda()
        y = y.cuda()

    # prev = time.time()
    with torch.no_grad():
        y_hat = dec(x)
        x_hat = enc(y)
    # time_used.append(time.time()-prev)
    # print(torch.median(x), torch.median(y), torch.median(y_hat), torch.median(x_hat))
    # y_hat = (y_hat - y_hat.min()) / (y_hat.max() - y_hat.min())
    y_hat[y_hat > 1.] = 1.
    x_hat[x_hat > 1.] = 1.
    # print(x.min(), y.max(), y.min(), y.max(), y_hat.min(), y_hat.max())
    ed = time.time()
    time_used.append(ed - st)
    ssim_loss = ssim1(y_hat, y)
    ssim_loss_yx = ssim1(x_hat, x)
    # print(y_hat.cpu().numpy().dtype, y_hat.cpu().numpy().dtype.type)
    sys.stdout.write('\ninput %04d takes time of %.8f, MSELoss: %f, L1Loss: %f, SSIM: %f, PSNR: %f' % (
                        i + 1, ed - st, criterion(y_hat, y).item(), criterion1(y_hat, y).item(), ssim_loss, peak_signal_noise_ratio(y_hat.cpu().numpy(), y.cpu().numpy())))
    sys.stdout.write('\ninput %04d takes time of %.8f, MSELoss: %f, L1Loss: %f, SSIM: %f, PSNR: %f (y->x)' % (
                        i + 1, ed - st, criterion(x_hat, x).item(), criterion1(x_hat, x).item(), ssim_loss_yx, peak_signal_noise_ratio(x_hat.cpu().numpy(), x.cpu().numpy())))
    # sys.stdout.write('\nAverage MSELoss: %f, L1Loss: %f, SSIM: %f' % (avg_loss[0]/(i+1), avg_loss[1]/(i+1), avg_loss[2]/(i+1)))

    if opt.batch_size == 4:
        def convert4to1(data):
            # print(data.shape)
            data_row1 = torch.cat((data[0,:,:,:], data[1,:,:,:]), 2)
            data_row2 = torch.cat((data[2,:,:,:], data[3,:,:,:]), 2)
            # print(data_row1.shape)
            return torch.cat((data_row1, data_row2), 1)

        x = convert4to1(x).unsqueeze(0)
        y = convert4to1(y).unsqueeze(0)
        y_hat = convert4to1(y_hat).unsqueeze(0)
        x_hat = convert4to1(x_hat).unsqueeze(0)
    diff = torch.abs(y_hat - y)
    avg_loss[0].append(criterion(y_hat, y).item())
    avg_loss[1].append(criterion1(y_hat, y).item())
    avg_loss[2].append(ssim_loss.item())
    avg_loss[3].append(peak_signal_noise_ratio(y_hat.cpu().numpy(), y.cpu().numpy()))
    avg_loss_x[0].append(criterion(x_hat, x).item())
    avg_loss_x[1].append(criterion1(x_hat, x).item())
    avg_loss_x[2].append(ssim_loss_yx.item())
    avg_loss_x[3].append(peak_signal_noise_ratio(x_hat.cpu().numpy(), x.cpu().numpy()))
    if opt.test_epoch == -1:
        show_tif(x.squeeze(), y.squeeze(), y_hat.squeeze(), x_hat.squeeze(),
                       name='images/%s/version_%s/test/ft_%s' % (opt.dataset_name, opt.version, i), cuda=cuda, save_y=True)
        y_hat_expand = torch.cat((y_hat[:, 0, :, :], y_hat[:, 1, :, :], y_hat[:, 2, :, :]), 2)
        save_image(y_hat_expand, "images/%s/version_%s/test/ft_yhat_expand_%s.png" % (opt.dataset_name, opt.version, i), normalize=False)
    # elif i % 5 == 0:
    else:
        image_grid = show_tif(x.squeeze(), y.squeeze(), y_hat.squeeze(), x_hat.squeeze(),
                       name='images/%s/version_%s/test/%s' % (opt.dataset_name, opt.version, i), cuda=cuda, save_y=True, save_x=True)

# avg_loss[0] /= len(test_loader)
# avg_loss[1] /= len(test_loader)
# avg_loss[2] /= len(test_loader)
print('\nTime: ', np.mean(time_used))
with open(os.path.join('result', '%s_v%s_epoch_%s.txt' % (opt.dataset_name, opt.version, str(opt.test_epoch))), 'w') as f:
    f.write(str(opt))
    f.write('\nAverage MSELoss: %f, L1Loss: %f, SSIM: %f, PSNR: %f' % (np.mean(avg_loss[0]), np.mean(avg_loss[1]), np.mean(avg_loss[2]), np.mean(avg_loss[3])))
    f.write('\nStd: %f, %f, %f, %f' % (np.std(avg_loss[0]), np.std(avg_loss[1]), np.std(avg_loss[2]), np.std(avg_loss[3])))
    f.write('\nAverage MSELoss: %f, L1Loss: %f, SSIM: %f, PSNR: %f (y->x)\n' % (np.mean(avg_loss_x[0]), np.mean(avg_loss_x[1]), np.mean(avg_loss_x[2]), np.mean(avg_loss_x[3])))
    for i in avg_loss:
        f.write(str(i))
        f.write('\n')
    for i in avg_loss_x:
        f.write(str(i))
        f.write('\n')
sys.stdout.write('\nAverage MSELoss: %f, L1Loss: %f, SSIM: %f, PSNR: %f' % (np.mean(avg_loss[0]), np.mean(avg_loss[1]), np.mean(avg_loss[2]), np.mean(avg_loss[3])))
sys.stdout.write('\nAverage MSELoss: %f, L1Loss: %f, SSIM: %f, PSNR: %f (y->x)' % (np.mean(avg_loss_x[0]), np.mean(avg_loss_x[1]), np.mean(avg_loss_x[2]), np.mean(avg_loss_x[3])))