CapsPix2Pix/interpolation.py at master · CherBass/CapsPix2Pix · 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
import torch
import argparse
import json
import os
import time
from Capsule_Networks import capspix2pixG as NetG #change for different networks
from AxonDataset import AxonDataset
import torchvision.utils as vutils
from torch.autograd import Variable
from PIL import Image

def achieve_args():
    parse = argparse.ArgumentParser()
    parse.add_argument('--experiment_load', type=str, default='models/',
                       help='model to load (path)')
    parse.add_argument('--model_load', type=str, default='ModelG_capspix2pix.pt',
                       help='model to load (name)')
    parse.add_argument('--data_load', type=str, default='crops256_inter',
                       help='data labels for interpolation')
    parse.add_argument('--norm_output', type=bool, default=True,
                       help='whether to normalise the features')

    args = parse.parse_args()
    return args


if __name__ == '__main__':

    args = achieve_args()
    args = vars(args)
    args['cuda'] = torch.cuda.is_available()

    # Setting parameters
    timestr = time.strftime("%H%M%S")

    __location__ = os.path.realpath(
        os.path.join(os.getcwd(), os.path.dirname(__file__)))

    experiment = args['experiment_load']
    directory = experiment
    path = os.path.join(__location__, directory)

    with open(path+'/parameters.json') as file:
        params = json.load(file)

    args.update(params)
    args['val_batch_size'] = 8
    args['dynamic_routing'] = 'full'

    num_reps = 100

    if not os.path.exists(path + '/inter'):
        os.makedirs(path + '/inter')

    netG = NetG(args)
    if args['cuda']:
        netG = netG.cuda()

    netG.load_state_dict(torch.load(path+'/'+args['model_load']))

    # We use dataLoader to get the images of the training set batch by batch
    axon_dataset_val = AxonDataset(data_name=args['data_load'])
    valDataloader = torch.utils.data.DataLoader(axon_dataset_val, batch_size=args['val_batch_size'],
                                                shuffle=True)
    netG.train()

    for n in range(num_reps):
        val_data = torch.FloatTensor(args['val_batch_size'], 1, args['image_size'], args['image_size'])
        val_label = torch.FloatTensor(args['val_batch_size'], 1, args['image_size'], args['image_size'])

        if args['cuda']:
            val_data, val_label = val_data.cuda(), val_label.cuda()

        #data
        val_iter = iter(valDataloader)
        data_val = val_iter.next()
        val_data_cpu, val_label_cpu = data_val
        if args['cuda']:
            val_label_cpu, val_data_cpu = val_label_cpu.cuda(), val_data_cpu.cuda()

        val_label.resize_as_(val_label_cpu).copy_(val_label_cpu)
        # val_label[0, 0, :, :] = torch.zeros((1, args['val_image_size'], args['val_image_size']))
        val_data.resize_as_(val_data_cpu).copy_(val_data_cpu)
        # val_data[0, 0, :, :] = torch.zeros((1, args['val_image_size'], args['val_image_size']))

        for i in range(val_label.size()[0]):
            val_label[i,:,:,:] = val_label[1,:,:,:]
            val_data[i,:,:,:] = val_data[1,:,:,:]

        #vutils.save_image(val_data[0], '%s/syn_target_inter_%03d.png' % (path, n), normalize=True)
        vutils.save_image(val_label[0], '%s/inter/syn_input_inter_%03d.png' % (path, n), normalize=True)
        val_data, val_label = Variable(val_data), Variable(val_label)

        # run generator on validation images
        if not (args['noise_source'] == 'dropout'):
            netG.eval()

        if args['noise_source'] == 'input':
            val_noise = torch.randn(val_label.size()[0], args['noise_size'])
        elif (args['noise_source'] == 'broadcast'):
            val_noise = torch.randn(val_label.size()[0], args['noise_size'], 1)
            num_copies = args['val_image_size'] // args['noise_size']
            if args['val_image_size'] % args['noise_size'] == 0:
                val_noise = val_noise.repeat(1, num_copies, args['val_image_size'])  # specifies number of copies
                val_noise = val_noise.unsqueeze(1)
            else:
                print('noise size is indivisible by image size')
        elif (args['noise_source'] == 'broadcast_conv'):
            val_noise = torch.randn(val_label.size()[0], args['noise_size'], 1, 1)
            val_noise = val_noise.repeat(1, 1, args['val_image_size'],
                                         args['val_image_size'])  # specifies number of copies
        else:
            val_noise = torch.zeros(0)

        if args['cuda']:
            val_noise = val_noise.cuda()
            val_noise = Variable(val_noise)

        args['state'] = 'val'

        fake_val = torch.zeros(val_label.size(0), 1,  args['image_size'], args['image_size'])
        for n_z in range(0, fake_val.size(0)):
            fake_val_temp, _ = netG(val_label[n_z].unsqueeze(0), val_noise[n_z].unsqueeze(0), args)
            fake_val[n_z], _ = fake_val_temp.data, fake_val_temp.data

        # fake_val, _ = netG(val_label, val_noise, args)
        vutils.save_image(fake_val.data, '%s/inter/syn_samples_random_%03d.png' % (path, n), normalize=True)

        # image transition
        vec_ind=1
        z1 = torch.randn(1, args['noise_size'])
        # z1 = torch.FloatTensor(1, args['noise_size']).normal_(0,1)
        temp = torch.FloatTensor(1, args['noise_size'])
        temp.copy_(z1)
        z2 =  torch.randn(1, args['noise_size'])
        dz = (z2 - z1) / args['val_batch_size']
        z = torch.FloatTensor(val_label.size()[0], args['noise_size'])
        for i in range(val_label.size()[0]):
            temp[:, :] = z1[:, :] + i * dz[:, :]
            z[i, :] = temp

        if args['cuda']:
            z = z.cuda()

        z_out = torch.zeros(z.size(0), 1,  args['image_size'], args['image_size'])
        x_out = torch.zeros(z.size(0), 16, args['image_size'], args['image_size'])
        x_out_ch = torch.zeros(16, args['image_size'], args['image_size'])

        for n_z in range(0, x_out.size(0)):
            z_out_temp, x_out_temp = netG(val_label[n_z].unsqueeze(0), z[n_z].unsqueeze(0), args)
            z_out[n_z], x_out[n_z] = z_out_temp.data, x_out_temp.data


        for n_x in range(0, x_out.size(1)):
            x_out_temp = x_out[:,n_x,:,:]
            x_out_temp = x_out_temp.unsqueeze(1)
            if args['norm_output']:
                x_temp = x_out_temp
                x_temp = (x_temp - x_temp.min()) / (x_temp.max() - x_temp.min())
                x_out_temp.data = x_temp.data
            x_out_ch[n_x] = x_out_temp.data[0,0,:,:]

            vutils.save_image(x_out_temp.data, '%s/inter/val_caps_interpolation_n_%03d_x_%03d.png' % (path, n, n_x), normalize=True, nrow=args['val_batch_size'])
        x_out_ch = x_out_ch.unsqueeze(1)
        vutils.save_image(z_out.data, '%s/inter/val_caps_interpolation_%03d.png' % (path, n), normalize=True, nrow=args['val_batch_size'])
        vutils.save_image(x_out_ch.data, '%s/inter/val_caps_%03d.png' % (path, n), normalize=True, nrow=16//8)


    print('finished')