- 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
- 🍖 原作者:K同学啊 | 接辅导、项目定制
- 🚀 文章来源:K同学的学习圈子
一、理论基础
1.什么是条件生成对抗网络(Conditional Generative Adversarial Network,CGAN)
CGAN是一种生成对抗网络(GAN)的变体,它引入了条件信息,以指导生成器生成具有特定属性的逼真样本,从而在生成数据的过程中更具有控制性和可解释性。
在CGAN中,生成器和判别器仍然是两个关键组件,生成器负责根据给定的条件信息生成合成数据,而判别器则努力区分真实数据和生成数据。与普通的GAN相比,CGAN在模型架构和训练过程中引入了额外的条件信息,这一信息可以是任何有助于生成具有特定属性的数据的附加输入,比如类别标签、文本描述等。
关键特点包括:
-
条件输入: CGAN的生成器和判别器接收额外的条件信息,这有助于控制生成过程和生成具有特定属性的样本。
-
有监督学习: 通过提供真实样本的标签作为条件,CGAN可以进行有监督学习,从而生成符合特定标签的样本。
-
灵活性和可解释性: 条件信息的引入使得生成数据更具有语义解释性和可控性,可以用于许多实际应用中需要特定属性的数据生成任务。
-
多样性控制: 通过改变条件信息,可以灵活地控制生成器输出的多样性,从而获得不同风格或属性的样本。
CGAN通过引入条件信息,使得生成过程更加可控和可解释,因此在诸如图像生成、图像编辑等任务中具有广泛的应用前景。
2. CGAN原理
CGAN(Conditional Generative Adversarial Network)是对GAN的一种扩展,引入了条件输入,使生成器可以受到额外信息的指导以生成特定类别的数据。
与传统的GAN不同,CGAN包括两个关键组件:生成器和判别器。生成器接收两部分输入:随机噪声向量和条件信息,如类别标签。它的任务是生成与给定条件相对应的真实数据样本。生成器通常由多层卷积和反卷积层构成,以将输入噪声和条件信息转换为逼真的输出。
判别器的作用是评估生成的数据样本是否真实。与DCGAN相似,判别器由多层卷积层组成,它接收生成器生成的数据以及相应的条件信息作为输入,并尝试将它们与真实数据区分开来。
在训练过程中,生成器和判别器交替进行学习。生成器试图生成与条件匹配的逼真数据,同时判别器试图准确识别生成的数据是否为真实。通过这种竞争和合作的过程,生成器逐渐提高生成数据的质量,直到可以生成与条件信息相匹配的高质量样本。CGAN的引入使得生成器可以更加精确地控制生成的数据,使其适用于更多的应用场景,如图像生成、文本生成等。
3.CGAN与GAN相同点与不同点
CGAN(Conditional Generative Adversarial Network)和GAN(Generative Adversarial Network)在基本原理上有相似之处,都是通过两个相互对抗的网络来进行无监督学习,以生成高质量的新数据。然而,它们在以下方面存在一些显著的异同:
相同点:
-
基本原理相同:CGAN和GAN都采用生成对抗网络(Generative Adversarial Networks)的基本原理,即通过生成器和判别器两个子网络的对抗训练来提高生成数据的质量。
-
判别器结构相同:在CGAN和GAN中,判别器结构都是相同的,通常都采用卷积神经网络(CNN)进行特征提取和分类。
不同点:
-
条件输入:CGAN引入了条件输入的概念,即生成器接收额外的条件信息,如类别标签,以指导生成过程。而GAN通常没有条件输入,生成器只接收随机噪声作为输入。
-
网络结构不同:CGAN和GAN的网络结构有所不同。CGAN的生成器和判别器可能需要额外的输入层来处理条件信息,而GAN通常不需要。
-
生成器和判别器输入输出不同:在CGAN中,生成器接收随机噪声和条件信息作为输入,并生成与条件相对应的输出。而GAN的生成器只接收随机噪声作为输入,并生成数据样本。
-
训练方法不同:CGAN的训练方法涉及到条件信息的传递和利用,而GAN通常不涉及这一点。
CGAN在GAN的基础上引入了条件输入的概念,使得生成器可以更加精确地控制生成的数据,适用于更多的应用场景,如图像生成、文本生成等,而GAN则更为简单直接,适用于一般的生成任务。
二、代码实现
2.1 准备工作
2.1.1导入依赖包
import torch
import torch.nn as nn
import numpy as np
import torch.optim as optim
from torchvision import datasets,transforms
from torch.autograd import Variable
from torchvision.utils import save_image
from torchvision.utils import make_grid
from torch.utils.tensorboard import SummaryWriter
from torchsummary import summary
import matplotlib.pyplot as plt
torch.manual_seed(1)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
batch_size = 128
<torch._C.Generator at 0x1e0fbfde670>
2.1.2 数据导入
train_transform = transforms.Compose([
transforms.Resize(int(128* 1.12)), ## 图片放大1.12倍
transforms.RandomCrop((128, 128)), ## 随机裁剪成原来的大小
transforms.ToTensor(),
transforms.Normalize([0.5,0.5,0.5], [0.5,0.5,0.5])])
train_dataset = datasets.ImageFolder(root='E:/BaiduNetdiskDownload/GAN-Data/rps/', transform=train_transform)
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
shuffle=True,
num_workers=6)
2.1.3 可视化
def show_images(images):
fig, ax = plt.subplots(figsize=(20, 20))
ax.set_xticks([]); ax.set_yticks([])
ax.imshow(make_grid(images.detach(), nrow=22).permute(1, 2, 0))
def show_batch(dl):
for images, _ in dl:
show_images(images)
break
show_batch(train_loader)
image_shape = (3, 128, 128)
image_dim = int(np.prod(image_shape))
latent_dim = 100
n_classes = 3
embedding_dim = 100
2.2 模型构建
2.2.1 初始化
# 自定义权重初始化函数,用于初始化生成器和判别器的权重
def weights_init(m):
# 获取当前层的类名
classname = m.__class__.__name__
# 如果当前层是卷积层(类名中包含 'Conv' )
if classname.find('Conv') != -1:
# 使用正态分布随机初始化权重,均值为0,标准差为0.02
torch.nn.init.normal_(m.weight, 0.0, 0.02)
# 如果当前层是批归一化层(类名中包含 'BatchNorm' )
elif classname.find('BatchNorm') != -1:
# 使用正态分布随机初始化权重,均值为1,标准差为0.02
torch.nn.init.normal_(m.weight, 1.0, 0.02)
# 将偏置项初始化为全零
torch.nn.init.zeros_(m.bias)
2.2.2 生成器
class Generator(nn.Module):
def __init__(self):
super(Generator, self).__init__()
# 定义条件标签的生成器部分,用于将标签映射到嵌入空间中
# n_classes:条件标签的总数
# embedding_dim:嵌入空间的维度
self.label_conditioned_generator = nn.Sequential(
nn.Embedding(n_classes, embedding_dim), # 使用Embedding层将条件标签映射为稠密向量
nn.Linear(embedding_dim, 16) # 使用线性层将稠密向量转换为更高维度
)
# 定义潜在向量的生成器部分,用于将噪声向量映射到图像空间中
# latent_dim:潜在向量的维度
self.latent = nn.Sequential(
nn.Linear(latent_dim, 4*4*512), # 使用线性层将潜在向量转换为更高维度
nn.LeakyReLU(0.2, inplace=True) # 使用LeakyReLU激活函数进行非线性映射
)
# 定义生成器的主要结构,将条件标签和潜在向量合并成生成的图像
self.model = nn.Sequential(
# 反卷积层1:将合并后的向量映射为64x8x8的特征图
nn.ConvTranspose2d(513, 64*8, 4, 2, 1, bias=False),
nn.BatchNorm2d(64*8, momentum=0.1, eps=0.8), # 批标准化
nn.ReLU(True), # ReLU激活函数
# 反卷积层2:将64x8x8的特征图映射为64x4x4的特征图
nn.ConvTranspose2d(64*8, 64*4, 4, 2, 1, bias=False),
nn.BatchNorm2d(64*4, momentum=0.1, eps=0.8),
nn.ReLU(True),
# 反卷积层3:将64x4x4的特征图映射为64x2x2的特征图
nn.ConvTranspose2d(64*4, 64*2, 4, 2, 1, bias=False),
nn.BatchNorm2d(64*2, momentum=0.1, eps=0.8),
nn.ReLU(True),
# 反卷积层4:将64x2x2的特征图映射为64x1x1的特征图
nn.ConvTranspose2d(64*2, 64*1, 4, 2, 1, bias=False),
nn.BatchNorm2d(64*1, momentum=0.1, eps=0.8),
nn.ReLU(True),
# 反卷积层5:将64x1x1的特征图映射为3x64x64的RGB图像
nn.ConvTranspose2d(64*1, 3, 4, 2, 1, bias=False),
nn.Tanh() # 使用Tanh激活函数将生成的图像像素值映射到[-1, 1]范围内
)
def forward(self, inputs):
noise_vector, label = inputs
# 通过条件标签生成器将标签映射为嵌入向量
label_output = self.label_conditioned_generator(label)
# 将嵌入向量的形状变为(batch_size, 1, 4, 4),以便与潜在向量进行合并
label_output = label_output.view(-1, 1, 4, 4)
# 通过潜在向量生成器将噪声向量映射为潜在向量
latent_output = self.latent(noise_vector)
# 将潜在向量的形状变为(batch_size, 512, 4, 4),以便与条件标签进行合并
latent_output = latent_output.view(-1, 512, 4, 4)
# 将条件标签和潜在向量在通道维度上进行合并,得到合并后的特征图
concat = torch.cat((latent_output, label_output), dim=1)
# 通过生成器的主要结构将合并后的特征图生成为RGB图像
image = self.model(concat)
return image
generator = Generator().to(device)
generator.apply(weights_init)
print(generator)
结构如下:
Generator(
(label_conditioned_generator): Sequential(
(0): Embedding(3, 100)
(1): Linear(in_features=100, out_features=16, bias=True)
)
(latent): Sequential(
(0): Linear(in_features=100, out_features=8192, bias=True)
(1): LeakyReLU(negative_slope=0.2, inplace=True)
)
(model): Sequential(
(0): ConvTranspose2d(513, 512, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
(6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(7): BatchNorm2d(128, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(8): ReLU(inplace=True)
(9): ConvTranspose2d(128, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(10): BatchNorm2d(64, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(11): ReLU(inplace=True)
(12): ConvTranspose2d(64, 3, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(13): Tanh()
)
)
from torchinfo import summary
summary(generator)
结构如下:
=================================================================
Layer (type:depth-idx) Param #
=================================================================
Generator –
├─Sequential: 1-1 –
│ └─Embedding: 2-1 300
│ └─Linear: 2-2 1,616
├─Sequential: 1-2 –
│ └─Linear: 2-3 827,392
│ └─LeakyReLU: 2-4 –
├─Sequential: 1-3 –
│ └─ConvTranspose2d: 2-5 4,202,496
│ └─BatchNorm2d: 2-6 1,024
│ └─ReLU: 2-7 –
│ └─ConvTranspose2d: 2-8 2,097,152
│ └─BatchNorm2d: 2-9 512
│ └─ReLU: 2-10 –
│ └─ConvTranspose2d: 2-11 524,288
│ └─BatchNorm2d: 2-12 256
│ └─ReLU: 2-13 –
│ └─ConvTranspose2d: 2-14 131,072
│ └─BatchNorm2d: 2-15 128
│ └─ReLU: 2-16 –
│ └─ConvTranspose2d: 2-17 3,072
│ └─Tanh: 2-18 –
=================================================================
Total params: 7,789,308
Trainable params: 7,789,308
Non-trainable params: 0
=================================================================
a = torch.ones(100)
b = torch.ones(1)
b = b.long()
a = a.to(device)
b = b.to(device)
2.2.3 判别器
import torch
import torch.nn as nn
class Discriminator(nn.Module):
def __init__(self):
super(Discriminator, self).__init__()
# 定义一个条件标签的嵌入层,用于将类别标签转换为特征向量
self.label_condition_disc = nn.Sequential(
nn.Embedding(n_classes, embedding_dim), # 嵌入层将类别标签编码为固定长度的向量
nn.Linear(embedding_dim, 3*128*128) # 线性层将嵌入的向量转换为与图像尺寸相匹配的特征张量
)
# 定义主要的鉴别器模型
self.model = nn.Sequential(
nn.Conv2d(6, 64, 4, 2, 1, bias=False), # 输入通道为6(包含图像和标签的通道数),输出通道为64,4x4的卷积核,步长为2,padding为1
nn.LeakyReLU(0.2, inplace=True), # LeakyReLU激活函数,带有负斜率,增加模型对输入中的负值的感知能力
nn.Conv2d(64, 64*2, 4, 3, 2, bias=False), # 输入通道为64,输出通道为64*2,4x4的卷积核,步长为3,padding为2
nn.BatchNorm2d(64*2, momentum=0.1, eps=0.8), # 批量归一化层,有利于训练稳定性和收敛速度
nn.LeakyReLU(0.2, inplace=True),
nn.Conv2d(64*2, 64*4, 4, 3, 2, bias=False), # 输入通道为64*2,输出通道为64*4,4x4的卷积核,步长为3,padding为2
nn.BatchNorm2d(64*4, momentum=0.1, eps=0.8),
nn.LeakyReLU(0.2, inplace=True),
nn.Conv2d(64*4, 64*8, 4, 3, 2, bias=False), # 输入通道为64*4,输出通道为64*8,4x4的卷积核,步长为3,padding为2
nn.BatchNorm2d(64*8, momentum=0.1, eps=0.8),
nn.LeakyReLU(0.2, inplace=True),
nn.Flatten(), # 将特征图展平为一维向量,用于后续全连接层处理
nn.Dropout(0.4), # 随机失活层,用于减少过拟合风险
nn.Linear(4608, 1), # 全连接层,将特征向量映射到输出维度为1的向量
nn.Sigmoid() # Sigmoid激活函数,用于输出范围限制在0到1之间的概率值
)
def forward(self, inputs):
img, label = inputs
# 将类别标签转换为特征向量
label_output = self.label_condition_disc(label)
# 重塑特征向量为与图像尺寸相匹配的特征张量
label_output = label_output.view(-1, 3, 128, 128)
# 将图像特征和标签特征拼接在一起作为鉴别器的输入
concat = torch.cat((img, label_output), dim=1)
# 将拼接后的输入通过鉴别器模型进行前向传播,得到输出结果
output = self.model(concat)
return output
discriminator = Discriminator().to(device)
discriminator.apply(weights_init)
print(discriminator)
结构如下:
Discriminator(
(label_condition_disc): Sequential(
(0): Embedding(3, 100)
(1): Linear(in_features=100, out_features=49152, bias=True)
)
(model): Sequential(
(0): Conv2d(6, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(1): LeakyReLU(negative_slope=0.2, inplace=True)
(2): Conv2d(64, 128, kernel_size=(4, 4), stride=(3, 3), padding=(2, 2), bias=False)
(3): BatchNorm2d(128, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(4): LeakyReLU(negative_slope=0.2, inplace=True)
(5): Conv2d(128, 256, kernel_size=(4, 4), stride=(3, 3), padding=(2, 2), bias=False)
(6): BatchNorm2d(256, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(7): LeakyReLU(negative_slope=0.2, inplace=True)
(8): Conv2d(256, 512, kernel_size=(4, 4), stride=(3, 3), padding=(2, 2), bias=False)
(9): BatchNorm2d(512, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(10): LeakyReLU(negative_slope=0.2, inplace=True)
(11): Flatten(start_dim=1, end_dim=-1)
(12): Dropout(p=0.4, inplace=False)
(13): Linear(in_features=4608, out_features=1, bias=True)
(14): Sigmoid()
)
)
summary(discriminator)
结构如下:
=================================================================
Layer (type:depth-idx) Param #
=================================================================
Discriminator –
├─Sequential: 1-1 –
│ └─Embedding: 2-1 300
│ └─Linear: 2-2 4,964,352
├─Sequential: 1-2 –
│ └─Conv2d: 2-3 6,144
│ └─LeakyReLU: 2-4 –
│ └─Conv2d: 2-5 131,072
│ └─BatchNorm2d: 2-6 256
│ └─LeakyReLU: 2-7 –
│ └─Conv2d: 2-8 524,288
│ └─BatchNorm2d: 2-9 512
│ └─LeakyReLU: 2-10 –
│ └─Conv2d: 2-11 2,097,152
│ └─BatchNorm2d: 2-12 1,024
│ └─LeakyReLU: 2-13 –
│ └─Flatten: 2-14 –
│ └─Dropout: 2-15 –
│ └─Linear: 2-16 4,609
│ └─Sigmoid: 2-17 –
=================================================================
Total params: 7,729,709
Trainable params: 7,729,709
Non-trainable params: 0
=================================================================
a = torch.ones(2,3,128,128)
b = torch.ones(2,1)
b = b.long()
a = a.to(device)
b = b.to(device)
c = discriminator((a,b))
c.size()
torch.Size([2, 1])
2.3 训练
2.3.1 定义损失函数
adversarial_loss = nn.BCELoss()
def generator_loss(fake_output, label):
gen_loss = adversarial_loss(fake_output, label)
return gen_loss
def discriminator_loss(output, label):
disc_loss = adversarial_loss(output, label)
return disc_loss
2.3.2. 定义优化器
learning_rate = 0.0002
G_optimizer = optim.Adam(generator.parameters(), lr = learning_rate, betas=(0.5, 0.999))
D_optimizer = optim.Adam(discriminator.parameters(), lr = learning_rate, betas=(0.5, 0.999))
2.3.3 训练模型
# 设置训练的总轮数
num_epochs = 100
# 初始化用于存储每轮训练中判别器和生成器损失的列表
D_loss_plot, G_loss_plot = [], []
# 循环进行训练
for epoch in range(1, num_epochs + 1):
# 初始化每轮训练中判别器和生成器损失的临时列表
D_loss_list, G_loss_list = [], []
# 遍历训练数据加载器中的数据
for index, (real_images, labels) in enumerate(train_loader):
# 清空判别器的梯度缓存
D_optimizer.zero_grad()
# 将真实图像数据和标签转移到GPU(如果可用)
real_images = real_images.to(device)
labels = labels.to(device)
# 将标签的形状从一维向量转换为二维张量(用于后续计算)
labels = labels.unsqueeze(1).long()
# 创建真实目标和虚假目标的张量(用于判别器损失函数)
real_target = Variable(torch.ones(real_images.size(0), 1).to(device))
fake_target = Variable(torch.zeros(real_images.size(0), 1).to(device))
# 计算判别器对真实图像的损失
D_real_loss = discriminator_loss(discriminator((real_images, labels)), real_target)
# 从噪声向量中生成假图像(生成器的输入)
noise_vector = torch.randn(real_images.size(0), latent_dim, device=device)
noise_vector = noise_vector.to(device)
generated_image = generator((noise_vector, labels))
# 计算判别器对假图像的损失(注意detach()函数用于分离生成器梯度计算图)
output = discriminator((generated_image.detach(), labels))
D_fake_loss = discriminator_loss(output, fake_target)
# 计算判别器总体损失(真实图像损失和假图像损失的平均值)
D_total_loss = (D_real_loss + D_fake_loss) / 2
D_loss_list.append(D_total_loss)
# 反向传播更新判别器的参数
D_total_loss.backward()
D_optimizer.step()
# 清空生成器的梯度缓存
G_optimizer.zero_grad()
# 计算生成器的损失
G_loss = generator_loss(discriminator((generated_image, labels)), real_target)
G_loss_list.append(G_loss)
# 反向传播更新生成器的参数
G_loss.backward()
G_optimizer.step()
# 打印当前轮次的判别器和生成器的平均损失
print('Epoch: [%d/%d]: D_loss: %.3f, G_loss: %.3f' % (
(epoch), num_epochs, torch.mean(torch.FloatTensor(D_loss_list)),
torch.mean(torch.FloatTensor(G_loss_list))))
# 将当前轮次的判别器和生成器的平均损失保存到列表中
D_loss_plot.append(torch.mean(torch.FloatTensor(D_loss_list)))
G_loss_plot.append(torch.mean(torch.FloatTensor(G_loss_list)))
if epoch%10 == 0:
# 将生成的假图像保存为图片文件
save_image(generated_image.data[:50], './sample_%d' % epoch + '.png', nrow=5, normalize=True)
# 将当前轮次的生成器和判别器的权重保存到文件
torch.save(generator.state_dict(), './generator_epoch_%d.pth' % (epoch))
torch.save(discriminator.state_dict(), './discriminator_epoch_%d.pth' % (epoch))
Epoch: [1/100]: D_loss: 0.282, G_loss: 1.860
Epoch: [2/100]: D_loss: 0.162, G_loss: 3.394
Epoch: [3/100]: D_loss: 0.233, G_loss: 3.330
Epoch: [4/100]: D_loss: 0.232, G_loss: 2.592
Epoch: [5/100]: D_loss: 0.254, G_loss: 2.471
Epoch: [6/100]: D_loss: 0.342, G_loss: 2.857
Epoch: [7/100]: D_loss: 0.400, G_loss: 1.995
Epoch: [8/100]: D_loss: 0.251, G_loss: 2.119
Epoch: [9/100]: D_loss: 0.402, G_loss: 2.091
Epoch: [10/100]: D_loss: 0.357, G_loss: 2.111
Epoch: [11/100]: D_loss: 0.563, G_loss: 1.845
Epoch: [12/100]: D_loss: 0.584, G_loss: 1.486
Epoch: [13/100]: D_loss: 0.591, G_loss: 1.420
Epoch: [14/100]: D_loss: 0.543, G_loss: 1.378
Epoch: [15/100]: D_loss: 0.568, G_loss: 1.304
Epoch: [16/100]: D_loss: 0.513, G_loss: 1.231
Epoch: [17/100]: D_loss: 0.550, G_loss: 1.393
Epoch: [18/100]: D_loss: 0.577, G_loss: 1.282
Epoch: [19/100]: D_loss: 0.549, G_loss: 1.245
Epoch: [20/100]: D_loss: 0.564, G_loss: 1.231
Epoch: [21/100]: D_loss: 0.543, G_loss: 1.382
Epoch: [22/100]: D_loss: 0.496, G_loss: 1.503
Epoch: [23/100]: D_loss: 0.502, G_loss: 1.451
Epoch: [24/100]: D_loss: 0.521, G_loss: 1.580
Epoch: [25/100]: D_loss: 0.476, G_loss: 1.565
Epoch: [26/100]: D_loss: 0.447, G_loss: 1.560
Epoch: [27/100]: D_loss: 0.494, G_loss: 1.677
Epoch: [28/100]: D_loss: 0.429, G_loss: 1.600
Epoch: [29/100]: D_loss: 0.466, G_loss: 1.709
Epoch: [30/100]: D_loss: 0.500, G_loss: 1.694
Epoch: [31/100]: D_loss: 0.479, G_loss: 1.549
Epoch: [32/100]: D_loss: 0.456, G_loss: 1.483
Epoch: [33/100]: D_loss: 0.513, G_loss: 1.393
Epoch: [34/100]: D_loss: 0.514, G_loss: 1.372
Epoch: [35/100]: D_loss: 0.475, G_loss: 1.332
Epoch: [36/100]: D_loss: 0.497, G_loss: 1.351
Epoch: [37/100]: D_loss: 0.477, G_loss: 1.337
Epoch: [38/100]: D_loss: 0.476, G_loss: 1.375
Epoch: [39/100]: D_loss: 0.488, G_loss: 1.435
Epoch: [40/100]: D_loss: 0.448, G_loss: 1.490
Epoch: [41/100]: D_loss: 0.455, G_loss: 1.494
Epoch: [42/100]: D_loss: 0.439, G_loss: 1.512
Epoch: [43/100]: D_loss: 0.445, G_loss: 1.540
Epoch: [44/100]: D_loss: 0.430, G_loss: 1.586
Epoch: [45/100]: D_loss: 0.442, G_loss: 1.692
Epoch: [46/100]: D_loss: 0.421, G_loss: 1.639
Epoch: [47/100]: D_loss: 0.422, G_loss: 1.734
Epoch: [48/100]: D_loss: 0.422, G_loss: 1.707
Epoch: [49/100]: D_loss: 0.401, G_loss: 1.792
Epoch: [50/100]: D_loss: 0.408, G_loss: 1.778
Epoch: [51/100]: D_loss: 0.406, G_loss: 1.813
Epoch: [52/100]: D_loss: 0.408, G_loss: 1.871
Epoch: [53/100]: D_loss: 0.373, G_loss: 1.796
Epoch: [54/100]: D_loss: 0.376, G_loss: 1.851
Epoch: [55/100]: D_loss: 0.416, G_loss: 1.939
Epoch: [56/100]: D_loss: 0.360, G_loss: 1.969
Epoch: [57/100]: D_loss: 0.372, G_loss: 1.910
Epoch: [58/100]: D_loss: 0.526, G_loss: 2.313
Epoch: [59/100]: D_loss: 0.318, G_loss: 1.960
Epoch: [60/100]: D_loss: 0.316, G_loss: 1.937
Epoch: [61/100]: D_loss: 0.306, G_loss: 1.928
Epoch: [62/100]: D_loss: 0.353, G_loss: 1.993
Epoch: [63/100]: D_loss: 0.326, G_loss: 2.079
Epoch: [64/100]: D_loss: 0.413, G_loss: 2.154
Epoch: [65/100]: D_loss: 0.312, G_loss: 2.047
Epoch: [66/100]: D_loss: 0.320, G_loss: 2.096
Epoch: [67/100]: D_loss: 0.360, G_loss: 2.244
Epoch: [68/100]: D_loss: 0.312, G_loss: 2.142
Epoch: [69/100]: D_loss: 0.326, G_loss: 2.135
Epoch: [70/100]: D_loss: 0.272, G_loss: 2.186
Epoch: [71/100]: D_loss: 0.290, G_loss: 2.232
Epoch: [72/100]: D_loss: 0.420, G_loss: 2.423
Epoch: [73/100]: D_loss: 0.422, G_loss: 2.347
Epoch: [74/100]: D_loss: 0.278, G_loss: 2.255
Epoch: [75/100]: D_loss: 0.262, G_loss: 2.262
Epoch: [76/100]: D_loss: 0.286, G_loss: 2.329
Epoch: [77/100]: D_loss: 0.289, G_loss: 2.416
Epoch: [78/100]: D_loss: 0.330, G_loss: 2.427
Epoch: [79/100]: D_loss: 0.337, G_loss: 2.522
Epoch: [80/100]: D_loss: 0.255, G_loss: 2.403
Epoch: [81/100]: D_loss: 0.289, G_loss: 2.468
Epoch: [82/100]: D_loss: 0.233, G_loss: 2.515
Epoch: [83/100]: D_loss: 0.274, G_loss: 2.547
Epoch: [84/100]: D_loss: 0.285, G_loss: 2.557
Epoch: [85/100]: D_loss: 0.244, G_loss: 2.536
Epoch: [86/100]: D_loss: 0.279, G_loss: 2.563
Epoch: [87/100]: D_loss: 0.394, G_loss: 2.705
Epoch: [88/100]: D_loss: 0.262, G_loss: 2.603
Epoch: [89/100]: D_loss: 0.254, G_loss: 2.613
Epoch: [90/100]: D_loss: 0.288, G_loss: 2.759
Epoch: [91/100]: D_loss: 0.275, G_loss: 2.576
Epoch: [92/100]: D_loss: 0.229, G_loss: 2.687
Epoch: [93/100]: D_loss: 0.232, G_loss: 2.682
Epoch: [94/100]: D_loss: 0.219, G_loss: 2.771
Epoch: [95/100]: D_loss: 0.588, G_loss: 3.084
Epoch: [96/100]: D_loss: 0.422, G_loss: 2.735
Epoch: [97/100]: D_loss: 0.254, G_loss: 2.705
Epoch: [98/100]: D_loss: 0.222, G_loss: 2.654
Epoch: [99/100]: D_loss: 0.229, G_loss: 2.757
Epoch: [100/100]: D_loss: 0.200, G_loss: 2.724
三、模型分析
generator.load_state_dict(torch.load('./training_weights/generator_epoch_100.pth'), strict=False)
generator.eval()
from numpy import asarray
from numpy.random import randn
from numpy.random import randint
from numpy import linspace
from matplotlib import pyplot
from matplotlib import gridspec
# 生成潜在空间的点,作为生成器的输入
def generate_latent_points(latent_dim, n_samples, n_classes=3):
# 从标准正态分布中生成潜在空间的点
x_input = randn(latent_dim * n_samples)
# 将生成的点整形成用于神经网络的输入的批量
z_input = x_input.reshape(n_samples, latent_dim)
return z_input
# 在两个潜在空间点之间进行均匀插值
def interpolate_points(p1, p2, n_steps=10):
# 在两个点之间进行插值,生成插值比率
ratios = linspace(0, 1, num=n_steps)
# 线性插值向量
vectors = list()
for ratio in ratios:
v = (1.0 - ratio) * p1 + ratio * p2
vectors.append(v)
return asarray(vectors)
# 生成两个潜在空间的点
pts = generate_latent_points(100, 2)
# 在两个潜在空间点之间进行插值
interpolated = interpolate_points(pts[0], pts[1])
# 将数据转换为torch张量并将其移至GPU(假设device已正确声明为GPU)
interpolated = torch.tensor(interpolated).to(device).type(torch.float32)
output = None
# 对于三个类别的循环,分别进行插值和生成图片
for label in range(3):
# 创建包含相同类别标签的张量
labels = torch.ones(10) * label
labels = labels.to(device)
labels = labels.unsqueeze(1).long()
print(labels.size())
# 使用生成器生成插值结果
predictions = generator((interpolated, labels))
predictions = predictions.permute(0,2,3,1)
pred = predictions.detach().cpu()
if output is None:
output = pred
else:
output = np.concatenate((output,pred))
nrow = 3
ncol = 10
fig = plt.figure(figsize=(15,4))
gs = gridspec.GridSpec(nrow, ncol)
k = 0
for i in range(nrow):
for j in range(ncol):
pred = (output[k, :, :, :] + 1 ) * 127.5
pred = np.array(pred)
ax= plt.subplot(gs[i,j])
ax.imshow(pred.astype(np.uint8))
ax.set_xticklabels([])
ax.set_yticklabels([])
ax.axis('off')
k += 1
plt.show()
Generator(
(label_conditioned_generator): Sequential(
(0): Embedding(3, 100)
(1): Linear(in_features=100, out_features=16, bias=True)
)
(latent): Sequential(
(0): Linear(in_features=100, out_features=8192, bias=True)
(1): LeakyReLU(negative_slope=0.2, inplace=True)
)
(model): Sequential(
(0): ConvTranspose2d(513, 512, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
(6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(7): BatchNorm2d(128, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(8): ReLU(inplace=True)
(9): ConvTranspose2d(128, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(10): BatchNorm2d(64, eps=0.8, momentum=0.1, affine=True, track_running_stats=True)
(11): ReLU(inplace=True)
(12): ConvTranspose2d(64, 3, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(13): Tanh()
)
)torch.Size([10, 1])
torch.Size([10, 1])
torch.Size([10, 1])
(30, 128, 128, 3)