1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

└─skip-ganomaly
    │  options.py
    │  train.py
    │  
    ├─experiments
    │      run_cifar.sh
    │      run_mnist.sh
    │      
    └─lib
        │  evaluate.py
        │  loss.py
        │  visualizer.py
        │  
        ├─data
        │      dataloader.py
        │      datasets.py
        │      
        └─models
                basemodel.py
                ganomaly.py
                networks.py
                skipganomaly.py
                __init__.py             

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

from options import Options
from lib.data.dataloader import load_data
from lib.models import load_model

##
def main():
    """ Training
    """
    opt = Options().parse()
    data = load_data(opt)
    model = load_model(opt, data)
    model.train()

if __name__ == '__main__':
    main()

 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

def load_data(opt):
    """ Load Data
    Args:
        opt ([type]): Argument Parser
    Returns:
        [type]: dataloader
    """
    # LOAD DATA SET
    if opt.dataroot == '':
        opt.dataroot = './data/{}'.format(opt.dataset)

    ## CIFAR
    if opt.dataset in ['cifar10']:
        transform = transforms.Compose([transforms.Resize(opt.isize),
                                        transforms.ToTensor(),
                                        transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])

        train_ds = CIFAR10(root='./data', train=True, download=True, transform=transform)
        valid_ds = CIFAR10(root='./data', train=False, download=True, transform=transform)
        train_ds, valid_ds = get_cifar_anomaly_dataset(train_ds, valid_ds, train_ds.class_to_idx[opt.abnormal_class])

    ## MNIST
    elif ...
    # FOLDER
    else:
        ...
        train_ds = ImageFolder(os.path.join(opt.dataroot, 'train'), transform)
        valid_ds = ImageFolder(os.path.join(opt.dataroot, 'test'), transform)
    ## DATALOADER
    train_dl = DataLoader(dataset=train_ds, batch_size=opt.batchsize, shuffle=True, drop_last=True)
    valid_dl = DataLoader(dataset=valid_ds, batch_size=opt.batchsize, shuffle=False, drop_last=False)
    return Data(train_dl, valid_dl)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18

Custom Dataset
├── test
│   ├── 0.normal
│   │   └── normal_tst_img_0.png
│   │   └── normal_tst_img_1.png
│   │   ...
│   │   └── normal_tst_img_n.png
│   ├── 1.abnormal
│   │   └── abnormal_tst_img_0.png
│   │   └── abnormal_tst_img_1.png
│   │   ...
│   │   └── abnormal_tst_img_m.png
├── train
│   ├── 0.normal
│   │   └── normal_tst_img_0.png
│   │   └── normal_tst_img_1.png
│   │   ...
│   │   └── normal_tst_img_t.png

1
2
3
4
5
6

def load_model(opt, dataloader):
    model_name = opt.model
    model_path = f"lib.models.{model_name}"
    model_lib  = importlib.import_module(model_path)
    model = getattr(model_lib, model_name.title())
    return model(opt, dataloader)

  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
225
226
227

class Skipganomaly(BaseModel):
    """GANomaly Class
    """
    @property
    def name(self): return 'skip-ganomaly'

    def __init__(self, opt, data=None):
        super(Skipganomaly, self).__init__(opt, data)
        ##

        # -- Misc attributes
        self.add_noise = True
        self.epoch = 0
        self.times = []
        self.total_steps = 0

        ##
        # Create and initialize networks.
        self.netg = define_G(self.opt, norm='batch', use_dropout=False, init_type='normal')
        self.netd = define_D(self.opt, norm='batch', use_sigmoid=False, init_type='normal')

        ##
        if self.opt.resume != '':
            print("\nLoading pre-trained networks.")
            self.opt.iter = torch.load(os.path.join(self.opt.resume, 'netG.pth'))['epoch']
            self.netg.load_state_dict(torch.load(os.path.join(self.opt.resume, 'netG.pth'))['state_dict'])
            self.netd.load_state_dict(torch.load(os.path.join(self.opt.resume, 'netD.pth'))['state_dict'])
            print("\tDone.\n")

        if self.opt.verbose:
            print(self.netg)
            print(self.netd)

        ##
        # Loss Functions
        self.l_adv = nn.BCELoss()
        self.l_con = nn.L1Loss()
        self.l_lat = l2_loss

        ##
        # Initialize input tensors.
        self.input = torch.empty(size=(self.opt.batchsize, 3, self.opt.isize, self.opt.isize), dtype=torch.float32, device=self.device)
        self.noise = torch.empty(size=(self.opt.batchsize, 3, self.opt.isize, self.opt.isize), dtype=torch.float32, device=self.device)
        self.label = torch.empty(size=(self.opt.batchsize,), dtype=torch.float32, device=self.device)
        self.gt = torch.empty(size=(opt.batchsize,), dtype=torch.long, device=self.device)
        self.fixed_input = torch.empty(size=(self.opt.batchsize, 3, self.opt.isize, self.opt.isize), dtype=torch.float32, device=self.device)
        self.real_label = torch.ones (size=(self.opt.batchsize,), dtype=torch.float32, device=self.device)
        self.fake_label = torch.zeros(size=(self.opt.batchsize,), dtype=torch.float32, device=self.device)

        ##
        # Setup optimizer
        if self.opt.isTrain:
            self.netg.train()
            self.netd.train()
            self.optimizers  = []
            self.optimizer_d = optim.Adam(self.netd.parameters(), lr=self.opt.lr, betas=(self.opt.beta1, 0.999))
            self.optimizer_g = optim.Adam(self.netg.parameters(), lr=self.opt.lr, betas=(self.opt.beta1, 0.999))
            self.optimizers.append(self.optimizer_d)
            self.optimizers.append(self.optimizer_g)
            self.schedulers = [get_scheduler(optimizer, opt) for optimizer in self.optimizers]

    def forward(self):
        self.forward_g()
        self.forward_d()

    def forward_g(self):
        """ Forward propagate through netG
        """
        self.fake = self.netg(self.input + self.noise)

    def forward_d(self):
        """ Forward propagate through netD
        """
        self.pred_real, self.feat_real = self.netd(self.input)
        self.pred_fake, self.feat_fake = self.netd(self.fake)

    def backward_g(self):
        """ Backpropagate netg
        """
        self.err_g_adv = self.opt.w_adv * self.l_adv(self.pred_fake, self.real_label)
        self.err_g_con = self.opt.w_con * self.l_con(self.fake, self.input)
        self.err_g_lat = self.opt.w_lat * self.l_lat(self.feat_fake, self.feat_real)

        self.err_g = self.err_g_adv + self.err_g_con + self.err_g_lat
        self.err_g.backward(retain_graph=True)

    def backward_d(self):
        # Fake
        pred_fake, _ = self.netd(self.fake.detach())
        self.err_d_fake = self.l_adv(pred_fake, self.fake_label)

        # Real
        # pred_real, feat_real = self.netd(self.input)
        self.err_d_real = self.l_adv(self.pred_real, self.real_label)

        # Combine losses.
        self.err_d = self.err_d_real + self.err_d_fake + self.err_g_lat
        self.err_d.backward(retain_graph=True)

    def update_netg(self):
        """ Update Generator Network.
        """       
        self.optimizer_g.zero_grad()
        self.backward_g()
        self.optimizer_g.step()

    def update_netd(self):
        """ Update Discriminator Network.
        """       
        self.optimizer_d.zero_grad()
        self.backward_d()
        self.optimizer_d.step()
        if self.err_d < 1e-5: self.reinit_d()
    ##
    def optimize_params(self):
        """ Optimize netD and netG  networks.
        """
        self.forward()
        self.update_netg()
        self.update_netd()

    ##
    def test(self, plot_hist=False):
        """ Test GANomaly model.

        Args:
            data ([type]): Dataloader for the test set

        Raises:
            IOError: Model weights not found.
        """
        with torch.no_grad():
            # Load the weights of netg and netd.
            if self.opt.load_weights:
                self.load_weights(is_best=True)

            self.opt.phase = 'test'

            scores = {}

            # Create big error tensor for the test set.
            self.an_scores = torch.zeros(size=(len(self.data.valid.dataset),), dtype=torch.float32, device=self.device)
            self.gt_labels = torch.zeros(size=(len(self.data.valid.dataset),), dtype=torch.long, device=self.device)
            self.features  = torch.zeros(size=(len(self.data.valid.dataset), self.opt.nz), dtype=torch.float32, device=self.device)

            print("   Testing %s" % self.name)
            self.times = []
            self.total_steps = 0
            epoch_iter = 0
            for i, data in enumerate(self.data.valid, 0):
                self.total_steps += self.opt.batchsize
                epoch_iter += self.opt.batchsize
                time_i = time.time()

                # Forward - Pass
                self.set_input(data)
                self.fake = self.netg(self.input)

                _, self.feat_real = self.netd(self.input)
                _, self.feat_fake = self.netd(self.fake)

                # Calculate the anomaly score.
                si = self.input.size()
                sz = self.feat_real.size()
                rec = (self.input - self.fake).view(si[0], si[1] * si[2] * si[3])
                lat = (self.feat_real - self.feat_fake).view(sz[0], sz[1] * sz[2] * sz[3])
                rec = torch.mean(torch.pow(rec, 2), dim=1)
                lat = torch.mean(torch.pow(lat, 2), dim=1)
                error = 0.9*rec + 0.1*lat

                time_o = time.time()

                self.an_scores[i*self.opt.batchsize: i*self.opt.batchsize + error.size(0)] = error.reshape(error.size(0))
                self.gt_labels[i*self.opt.batchsize: i*self.opt.batchsize + error.size(0)] = self.gt.reshape(error.size(0))

                self.times.append(time_o - time_i)

                # Save test images.
                if self.opt.save_test_images:
                    dst = os.path.join(self.opt.outf, self.opt.name, 'test', 'images')
                    if not os.path.isdir(dst): os.makedirs(dst)
                    real, fake, _ = self.get_current_images()
                    vutils.save_image(real, '%s/real_%03d.eps' % (dst, i+1), normalize=True)
                    vutils.save_image(fake, '%s/fake_%03d.eps' % (dst, i+1), normalize=True)

            # Measure inference time.
            self.times = np.array(self.times)
            self.times = np.mean(self.times[:100] * 1000)

            # Scale error vector between [0, 1]
            self.an_scores = (self.an_scores - torch.min(self.an_scores)) / \
                             (torch.max(self.an_scores) - torch.min(self.an_scores))
            auc = roc(self.gt_labels, self.an_scores)
            performance = OrderedDict([('Avg Run Time (ms/batch)', self.times), ('AUC', auc)])

            ##
            # PLOT HISTOGRAM
            if plot_hist:
                plt.ion()
                # Create data frame for scores and labels.
                scores['scores'] = self.an_scores
                scores['labels'] = self.gt_labels
                hist = pd.DataFrame.from_dict(scores)
                hist.to_csv("histogram.csv")

                # Filter normal and abnormal scores.
                abn_scr = hist.loc[hist.labels == 1]['scores']
                nrm_scr = hist.loc[hist.labels == 0]['scores']

                # Create figure and plot the distribution.
                # fig, ax = plt.subplots(figsize=(4,4));
                sns.distplot(nrm_scr, label=r'Normal Scores')
                sns.distplot(abn_scr, label=r'Abnormal Scores')

                plt.legend()
                plt.yticks([])
                plt.xlabel(r'Anomaly Scores')

            ##
            # PLOT PERFORMANCE
            if self.opt.display_id > 0 and self.opt.phase == 'test':
                counter_ratio = float(epoch_iter) / len(self.data.valid.dataset)
                self.visualizer.plot_performance(self.epoch, counter_ratio, performance)

            ##
            # RETURN
            return performance

1
2

self.netg = define_G(self.opt, norm='batch', use_dropout=False, init_type='normal')
self.netd = define_D(self.opt, norm='batch', use_sigmoid=False, init_type='normal')

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16

class UnetGenerator(nn.Module):
    def __init__(self, input_nc, output_nc, num_downs, ngf=64,
                 norm_layer=nn.BatchNorm2d, use_dropout=False):
        super(UnetGenerator, self).__init__()
        # construct unet structure
        unet_block = UnetSkipConnectionBlock(ngf * 8, ngf * 8, input_nc=None, submodule=None, norm_layer=norm_layer, innermost=True)
        for i in range(num_downs - 5):
            unet_block = UnetSkipConnectionBlock(ngf * 8, ngf * 8, input_nc=None, submodule=unet_block, norm_layer=norm_layer, use_dropout=use_dropout)
        unet_block = UnetSkipConnectionBlock(ngf * 4, ngf * 8, input_nc=None, submodule=unet_block, norm_layer=norm_layer)
        unet_block = UnetSkipConnectionBlock(ngf * 2, ngf * 4, input_nc=None, submodule=unet_block, norm_layer=norm_layer)
        unet_block = UnetSkipConnectionBlock(ngf, ngf * 2, input_nc=None, submodule=unet_block, norm_layer=norm_layer)
        unet_block = UnetSkipConnectionBlock(output_nc, ngf, input_nc=input_nc, submodule=unet_block, outermost=True, norm_layer=norm_layer)

        self.model = unet_block
    def forward(self, input):
        return self.model(input)

 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

class UnetSkipConnectionBlock(nn.Module):
    def __init__(self, outer_nc, inner_nc, input_nc=None,
                 submodule=None, outermost=False, innermost=False, norm_layer=nn.BatchNorm2d, use_dropout=False):
        super(UnetSkipConnectionBlock, self).__init__()
        self.outermost = outermost
        if type(norm_layer) == functools.partial:
            use_bias = norm_layer.func == nn.InstanceNorm2d
        else:
            use_bias = norm_layer == nn.InstanceNorm2d
        if input_nc is None:
            input_nc = outer_nc
        downconv = nn.Conv2d(input_nc, inner_nc, kernel_size=4,
                             stride=2, padding=1, bias=use_bias)
        downrelu = nn.LeakyReLU(0.2, True)
        downnorm = norm_layer(inner_nc)
        uprelu = nn.ReLU(True)
        upnorm = norm_layer(outer_nc)

        if outermost:
            upconv = nn.ConvTranspose2d(inner_nc * 2, outer_nc,
                                        kernel_size=4, stride=2,
                                        padding=1)
            down = [downconv]
            up = [uprelu, upconv, nn.Tanh()]
            model = down + [submodule] + up
        elif innermost:
            upconv = nn.ConvTranspose2d(inner_nc, outer_nc,
                                        kernel_size=4, stride=2,
                                        padding=1, bias=use_bias)
            down = [downrelu, downconv]
            up = [uprelu, upconv, upnorm]
            model = down + up
        else:
            upconv = nn.ConvTranspose2d(inner_nc * 2, outer_nc,
                                        kernel_size=4, stride=2,
                                        padding=1, bias=use_bias)
            down = [downrelu, downconv, downnorm]
            up = [uprelu, upconv, upnorm]

            if use_dropout:
                model = down + [submodule] + up + [nn.Dropout(0.5)]
            else:
                model = down + [submodule] + up

        self.model = nn.Sequential(*model)

    def forward(self, x):
        if self.outermost:
            return self.model(x)
        else:
            return torch.cat([x, self.model(x)], 1)

 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

def train_one_epoch(self):
        """ Train the model for one epoch.
        """

        self.netg.train()
        epoch_iter = 0
        for data in tqdm(self.data.train, leave=False, total=len(self.data.train)):
            self.total_steps += self.opt.batchsize
            epoch_iter += self.opt.batchsize

            self.set_input(data)
            self.optimize_params()

            if self.total_steps % self.opt.print_freq == 0:
                errors = self.get_errors()
                if self.opt.display:
                    counter_ratio = float(epoch_iter) / len(self.data.train.dataset)
                    self.visualizer.plot_current_errors(self.epoch, counter_ratio, errors)

            if self.total_steps % self.opt.save_image_freq == 0:
                reals, fakes, fixed = self.get_current_images()
                self.visualizer.save_current_images(self.epoch, reals, fakes, fixed)
                if self.opt.display:
                    self.visualizer.display_current_images(reals, fakes, fixed)

        print(">> Training model %s. Epoch %d/%d" % (self.name, self.epoch+1, self.opt.niter))

1
2
3
4
5
6

    def optimize_params(self):
        """ Optimize netD and netG  networks.
        """
        self.forward()
        self.update_netg()
        self.update_netd()