DeepDream-Image-Generator/train.py at master · akshit5565/DeepDream-Image-Generator · 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
# import torch
# import torch.nn as nn
# import torch.optim as optim
# import torchvision.datasets as dsets
# import torchvision.transforms as transforms #used for data preprocessing and augmentation
# import torchvision.utils as vutils # useful for visualizing and saving images
# from config import *
# from models import Generator, Discriminator

# results_dir = 'C:\\Users\\Akshit\\Desktop\\DeepDream-Image-Generator\\results'
# if not os.path.exists(results_dir):
#     os.makedirs(results_dir)

# real_samples_path = os.path.join(results_dir, 'real_samples.png')

# torch.manual_seed(manual_seed)

# transform = transforms.Compose([
#     transforms.Resize(image_size),
#     transforms.ToTensor(), # Converts the image to a PyTorch tensor and scales the pixel values to the range [0, 1]
#     transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5)) # Normalizes the tensor with mean 0.5 and standard deviation 0.5 for each of the three channels (R, G, B), scaling the pixel values to the range [-1, 1]
# ])

# dataset = dsets.CIFAR10(root='./data', download=True, transform=transform)
# dataloader = torch.utils.data.DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=2)

# device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

# generator = Generator(latent_dim).to(device)
# discriminator = Discriminator().to(device)

# criterion = nn.BCELoss() #binary cross entropy
# optimizer_G = optim.Adam(generator.parameters(), lr=lr, betas = (beta1,0.999))
# optimizer_D = optim.Adam(discriminator.parameters(), lr=lr, betas=(beta1,0.999))

# if __name__ == "__main__":
#     # Training loop
#     for epoch in range(num_epochs):
#         for i, data in enumerate(dataloader, 0):
#             discriminator.zero_grad()  # Clears the gradients of the discriminator

#             real_images, _ = data  # Unpacking real_images and corresponding labels, labels are ignored here
#             real_images = real_images.to(device)
#             batch_size = real_images.size(0)
#             label = torch.full((batch_size,), 1, device=device, dtype=torch.float)  # Creates a tensor of labels filled with ones (indicating real images) with the same batch size
#             output = discriminator(real_images).view(-1)  # Feeds the real images into the discriminator and flattens the output
#             errD_real = criterion(output, label)
#             errD_real.backward()  # Backpropagates the loss for real images
#             D_x = output.mean().item()  # Computes the mean of the discriminator's output for real images

#             noise = torch.randn(batch_size, latent_dim, 1, 1, device=device)  # Generates random noise
#             fake_images = generator(noise)  # Feeds noise into generator to produce fake images
#             label.fill_(0.0)  # Fills the label tensors with zeros indicating fake images
#             output = discriminator(fake_images.detach()).view(-1)  # Feeds fake images into discriminator and flattens the output
#             errD_fake = criterion(output, label)  # Computes BCE for fake images
#             errD_fake.backward()
#             D_G_z1 = output.mean().item()  # Computes mean of the discriminator's output for fake images

#             errD = errD_fake + errD_real  # Total discriminator loss
#             optimizer_D.step()  # Update discriminator's parameters

#             generator.zero_grad()  # Clear generator's gradients
#             label.fill_(1.0)  # Fills the label tensors with ones indicating real images
#             output = discriminator(fake_images).view(-1)
#             errG = criterion(output, label)
#             errG.backward()
#             D_G_z2 = output.mean().item()
#             optimizer_G.step()

#             if i % 100 == 0:
#                 print('[%d/%d][%d/%d] Loss_D: %.4f Loss_G: %.4f D(x): %.4f D(G(z)): %.4f / %.4f'
#                       % (epoch, num_epochs, i, len(dataloader),
#                          errD.item(), errG.item(), D_x, D_G_z1, D_G_z2))

#         # Save generated images
#         # Save generated images
#         if epoch == 0:
#             vutils.save_image(real_images, '%s/real_samples.png' % results_dir, normalize=True)
#         fake = generator(noise)
#         vutils.save_image(fake.detach(), '%s/fake_samples_epoch_%03d.png' % (results_dir, epoch), normalize=True)


#     # Save the trained model
#     torch.save(generator.state_dict(), 'generator.pth')


import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.datasets as dsets
import torchvision.transforms as transforms
import torchvision.utils as vutils
import os
from config import *
from models import Generator, Discriminator

results_dir = 'C:\\Users\\Akshit\\Desktop\\DeepDream-Image-Generator\\results'
if not os.path.exists(results_dir):
    os.makedirs(results_dir)

torch.manual_seed(manual_seed)

transform = transforms.Compose([
    transforms.Resize(image_size),
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])

# Load custom dataset from folder 'animeimages'
dataset = dsets.ImageFolder(root='animeimages', transform=transform)
dataloader = torch.utils.data.DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=2)

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

generator = Generator(latent_dim).to(device)
discriminator = Discriminator().to(device)

criterion = nn.BCELoss()
optimizer_G = optim.Adam(generator.parameters(), lr=lr, betas=(beta1, 0.999))
optimizer_D = optim.Adam(discriminator.parameters(), lr=lr, betas=(beta1, 0.999))

if __name__ == "__main__":
    for epoch in range(num_epochs):
        for i, data in enumerate(dataloader, 0):
            discriminator.zero_grad()

            real_images, _ = data
            real_images = real_images.to(device)
            batch_size = real_images.size(0)
            label = torch.full((batch_size,), 1, device=device, dtype=torch.float)
            output = discriminator(real_images).view(-1)
            errD_real = criterion(output, label)
            errD_real.backward()
            D_x = output.mean().item()

            noise = torch.randn(batch_size, latent_dim, 1, 1, device=device)
            fake_images = generator(noise)
            label.fill_(0.0)
            output = discriminator(fake_images.detach()).view(-1)
            errD_fake = criterion(output, label)
            errD_fake.backward()
            D_G_z1 = output.mean().item()

            errD = errD_fake + errD_real
            optimizer_D.step()

            generator.zero_grad()
            label.fill_(1.0)
            output = discriminator(fake_images).view(-1)
            errG = criterion(output, label)
            errG.backward()
            D_G_z2 = output.mean().item()
            optimizer_G.step()

            if i % 100 == 0:
                print('[%d/%d][%d/%d] Loss_D: %.4f Loss_G: %.4f D(x): %.4f D(G(z)): %.4f / %.4f'
                      % (epoch, num_epochs, i, len(dataloader),
                         errD.item(), errG.item(), D_x, D_G_z1, D_G_z2))

        # Save generated images
        if epoch == 0:
            vutils.save_image(real_images, '%s/real_samples.png' % results_dir, normalize=True)
        fake = generator(noise)
        vutils.save_image(fake.detach(), '%s/fake_samples_epoch_%03d.png' % (results_dir, epoch), normalize=True)

    # Save the trained model
    torch.save(generator.state_dict(), 'generator.pth')