第G7周：Semi-Supervised GAN 理论与实战

>- **🍨 本文为[🔗365天深度学习训练营](https://mp.weixin.qq.com/s/0dvHCaOoFnW8SCp3JpzKxg) 中的学习记录博客**

>- **🍖 原作者：[K同学啊](https://mtyjkh.blog.csdn.net/)**

复制代码

import argparse
import os
import numpy as np
import math

import torchvision.transforms as transforms
from torchvision.utils import save_image

from torch.utils.data import DataLoader
from torchvision import datasets
from torch.autograd import Variable

import torch.nn as nn
import torch.nn.functional as F
import torch

# 创建保存生成图像的文件夹
os.makedirs("images", exist_ok=True)

# 使用 argparse 解析命令行参数
parser = argparse.ArgumentParser()
parser.add_argument("--n_epochs", type=int, default=50, help="训练的轮数")
parser.add_argument("--batch_size", type=int, default=64, help="每个批次的样本数量")
parser.add_argument("--lr", type=float, default=0.0002, help="Adam 优化器的学习率")
parser.add_argument("--b1", type=float, default=0.5, help="Adam 优化器的第一个动量衰减参数")
parser.add_argument("--b2", type=float, default=0.999, help="Adam 优化器的第二个动量衰减参数")
parser.add_argument("--n_cpu", type=int, default=8, help="用于批次生成的 CPU 线程数")
parser.add_argument("--latent_dim", type=int, default=100, help="潜在空间的维度")
parser.add_argument("--num_classes", type=int, default=10, help="数据集的类别数")
parser.add_argument("--img_size", type=int, default=32, help="每个图像的尺寸（高度和宽度相等）")
parser.add_argument("--channels", type=int, default=1, help="图像的通道数（灰度图像通道数为 1）")
parser.add_argument("--sample_interval", type=int, default=400, help="图像采样间隔")
# opt = parser.parse_args()   #如果有这行代码，项目不适合用jupyter notebook运行
opt = parser.parse_args([])   #有这行代码，项目才适合在jupyter notebook运行，就是多加了'[]'
print(opt)

# 如果 GPU 可用，则使用 CUDA 加速
cuda = True if torch.cuda.is_available() else False

复制代码

Namespace(n_epochs=50, batch_size=64, lr=0.0002, b1=0.5, b2=0.999, n_cpu=8, latent_dim=100, num_classes=10, img_size=32, channels=1, sample_interval=400)

复制代码

def weights_init_normal(m):
    classname = m.__class__.__name__
    if classname.find("Conv") != -1:
        torch.nn.init.normal_(m.weight.data, 0.0, 0.02)
    elif classname.find("BatchNorm") != -1:
        torch.nn.init.normal_(m.weight.data, 1.0, 0.02)
        torch.nn.init.constant_(m.bias.data, 0.0)

import torch.nn as nn

class Generator(nn.Module):
    def __init__(self):
        super(Generator, self).__init__()

        # 创建一个标签嵌入层，用于将条件标签映射到潜在空间
        self.label_emb = nn.Embedding(opt.num_classes, opt.latent_dim)

        # 初始化图像尺寸，用于上采样之前
        self.init_size = opt.img_size // 4  # Initial size before upsampling

        # 第一个全连接层，将随机噪声映射到合适的维度
        self.l1 = nn.Sequential(nn.Linear(opt.latent_dim, 128 * self.init_size ** 2))

        # 生成器的卷积块
        self.conv_blocks = nn.Sequential(
            nn.BatchNorm2d(128),
            nn.Upsample(scale_factor=2),
            nn.Conv2d(128, 128, 3, stride=1, padding=1),
            nn.BatchNorm2d(128, 0.8),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Upsample(scale_factor=2),
            nn.Conv2d(128, 64, 3, stride=1, padding=1),
            nn.BatchNorm2d(64, 0.8),
            nn.LeakyReLU(0.2, inplace=True),
            nn.Conv2d(64, opt.channels, 3, stride=1, padding=1),
            nn.Tanh(),
        )

    def forward(self, noise):
        out = self.l1(noise)
        out = out.view(out.shape[0], 128, self.init_size, self.init_size)
        img = self.conv_blocks(out)
        return img

class Discriminator(nn.Module):
    def __init__(self):
        super(Discriminator, self).__init__()

        def discriminator_block(in_filters, out_filters, bn=True):
            """返回每个鉴别器块的层"""
            block = [nn.Conv2d(in_filters, out_filters, 3, 2, 1), nn.LeakyReLU(0.2, inplace=True), nn.Dropout2d(0.25)]
            if bn:
                block.append(nn.BatchNorm2d(out_filters, 0.8))
            return block

        # 鉴别器的卷积块
        self.conv_blocks = nn.Sequential(
            *discriminator_block(opt.channels, 16, bn=False),
            *discriminator_block(16, 32),
            *discriminator_block(32, 64),
            *discriminator_block(64, 128),
        )

        # 下采样图像的高度和宽度
        ds_size = opt.img_size // 2 ** 4

        # 输出层
        self.adv_layer = nn.Sequential(nn.Linear(128 * ds_size ** 2, 1), nn.Sigmoid())  # 用于鉴别真假的输出层
        self.aux_layer = nn.Sequential(nn.Linear(128 * ds_size ** 2, opt.num_classes + 1), nn.Softmax())  # 用于鉴别类别的输出层

    def forward(self, img):
        out = self.conv_blocks(img)
        out = out.view(out.shape[0], -1)
        validity = self.adv_layer(out)
        label = self.aux_layer(out)

        return validity, label

# 定义损失函数
adversarial_loss = torch.nn.BCELoss()  # 二元交叉熵损失，用于对抗训练
auxiliary_loss = torch.nn.CrossEntropyLoss()  # 交叉熵损失，用于辅助分类

# 初始化生成器和鉴别器
generator = Generator()  # 创建生成器实例
discriminator = Discriminator()  # 创建鉴别器实例

# 如果使用GPU，将模型和损失函数移至GPU上
if cuda:
    generator.cuda()
    discriminator.cuda()
    adversarial_loss.cuda()
    auxiliary_loss.cuda()

# 初始化模型权重
generator.apply(weights_init_normal)  # 初始化生成器的权重
discriminator.apply(weights_init_normal)  # 初始化鉴别器的权重

# 配置数据加载器
os.makedirs("./paper", exist_ok=True)  # 创建存储MNIST数据集的文件夹
dataloader = torch.utils.data.DataLoader(
    datasets.MNIST(
        "./paper",
        train=True,
        download=True,
        transform=transforms.Compose(
            [transforms.Resize(opt.img_size), transforms.ToTensor(), transforms.Normalize([0.5], [0.5])]
        ),
    ),
    batch_size=opt.batch_size,
    shuffle=True,
)

# 优化器
optimizer_G = torch.optim.Adam(generator.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))  # 生成器的优化器
optimizer_D = torch.optim.Adam(discriminator.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))  # 鉴别器的优化器

# 根据是否使用GPU选择数据类型
FloatTensor = torch.cuda.FloatTensor if cuda else torch.FloatTensor
LongTensor = torch.cuda.LongTensor if cuda else torch.LongTensor

复制代码

100%|█████████████████████████████████████████████████████████████████████████████| 9.91M/9.91M [00:02<00:00, 3.36MB/s]
100%|██████████████████████████████████████████████████████████████████████████████| 28.9k/28.9k [00:00<00:00, 206kB/s]
100%|█████████████████████████████████████████████████████████████████████████████| 1.65M/1.65M [00:00<00:00, 1.78MB/s]
100%|█████████████████████████████████████████████████████████████████████████████| 4.54k/4.54k [00:00<00:00, 5.50MB/s]

复制代码

for epoch in range(opt.n_epochs):
    for i, (imgs, labels) in enumerate(dataloader):

        batch_size = imgs.shape[0]

        # 定义对抗训练的标签
        valid = Variable(FloatTensor(batch_size, 1).fill_(1.0), requires_grad=False)  # 用于真实样本
        fake = Variable(FloatTensor(batch_size, 1).fill_(0.0), requires_grad=False)  # 用于生成样本
        fake_aux_gt = Variable(LongTensor(batch_size).fill_(opt.num_classes), requires_grad=False)  # 用于生成样本的类别标签

        # 配置输入数据
        real_imgs = Variable(imgs.type(FloatTensor))  # 真实图像
        labels = Variable(labels.type(LongTensor))  # 真实类别标签

        # -----------------
        #  训练生成器
        # -----------------

        optimizer_G.zero_grad()

        # 采样噪声和类别标签作为生成器的输入
        z = Variable(FloatTensor(np.random.normal(0, 1, (batch_size, opt.latent_dim))))

        # 生成一批图像
        gen_imgs = generator(z)

        # 计算生成器的损失，衡量生成器欺骗鉴别器的能力
        validity, _ = discriminator(gen_imgs)
        g_loss = adversarial_loss(validity, valid)

        g_loss.backward()
        optimizer_G.step()

        # ---------------------
        #  训练鉴别器
        # ---------------------

        optimizer_D.zero_grad()

        # 真实图像的损失
        real_pred, real_aux = discriminator(real_imgs)
        d_real_loss = (adversarial_loss(real_pred, valid) + auxiliary_loss(real_aux, labels)) / 2

        # 生成图像的损失
        fake_pred, fake_aux = discriminator(gen_imgs.detach())
        d_fake_loss = (adversarial_loss(fake_pred, fake) + auxiliary_loss(fake_aux, fake_aux_gt)) / 2

        # 总的鉴别器损失
        d_loss = (d_real_loss + d_fake_loss) / 2

        # 计算鉴别器准确率
        pred = np.concatenate([real_aux.data.cpu().numpy(), fake_aux.data.cpu().numpy()], axis=0)
        gt = np.concatenate([labels.data.cpu().numpy(), fake_aux_gt.data.cpu().numpy()], axis=0)
        d_acc = np.mean(np.argmax(pred, axis=1) == gt)

        d_loss.backward()
        optimizer_D.step()

        batches_done = epoch * len(dataloader) + i
        if batches_done % opt.sample_interval == 0:
            save_image(gen_imgs.data[:25], "images/%d.png" % batches_done, nrow=5, normalize=True)

    print(
        "[Epoch %d/%d] [Batch %d/%d] [D loss: %f, acc: %d%%] [G loss: %f]"
        % (epoch, opt.n_epochs, i, len(dataloader), d_loss.item(), 100 * d_acc, g_loss.item())
    )

复制代码

[Epoch 0/50] [Batch 937/938] [D loss: 1.362414, acc: 50%] [G loss: 0.707926]
[Epoch 1/50] [Batch 937/938] [D loss: 1.360016, acc: 50%] [G loss: 0.749867]
[Epoch 2/50] [Batch 937/938] [D loss: 1.345859, acc: 50%] [G loss: 0.901844]
[Epoch 3/50] [Batch 937/938] [D loss: 1.309943, acc: 50%] [G loss: 0.653252]
[Epoch 4/50] [Batch 937/938] [D loss: 1.340662, acc: 50%] [G loss: 0.742371]
[Epoch 5/50] [Batch 937/938] [D loss: 1.354004, acc: 50%] [G loss: 0.573718]
[Epoch 6/50] [Batch 937/938] [D loss: 1.326664, acc: 51%] [G loss: 0.729521]
[Epoch 7/50] [Batch 937/938] [D loss: 1.340284, acc: 48%] [G loss: 0.893304]
[Epoch 8/50] [Batch 937/938] [D loss: 1.297407, acc: 53%] [G loss: 0.512779]
[Epoch 9/50] [Batch 937/938] [D loss: 1.244959, acc: 51%] [G loss: 0.867025]
[Epoch 10/50] [Batch 937/938] [D loss: 1.409401, acc: 50%] [G loss: 0.626929]
[Epoch 11/50] [Batch 937/938] [D loss: 1.233004, acc: 53%] [G loss: 0.711049]
[Epoch 12/50] [Batch 937/938] [D loss: 1.339882, acc: 56%] [G loss: 0.841943]
[Epoch 13/50] [Batch 937/938] [D loss: 1.499406, acc: 50%] [G loss: 0.580588]
[Epoch 14/50] [Batch 937/938] [D loss: 1.242574, acc: 53%] [G loss: 1.458132]
[Epoch 15/50] [Batch 937/938] [D loss: 1.357433, acc: 54%] [G loss: 0.773089]
[Epoch 16/50] [Batch 937/938] [D loss: 1.319705, acc: 53%] [G loss: 0.728283]
[Epoch 17/50] [Batch 937/938] [D loss: 1.368955, acc: 46%] [G loss: 1.020731]
[Epoch 18/50] [Batch 937/938] [D loss: 1.383472, acc: 51%] [G loss: 0.933235]
[Epoch 19/50] [Batch 937/938] [D loss: 1.593254, acc: 54%] [G loss: 0.380877]
[Epoch 20/50] [Batch 937/938] [D loss: 1.242541, acc: 50%] [G loss: 0.761861]
[Epoch 21/50] [Batch 937/938] [D loss: 1.301212, acc: 62%] [G loss: 0.558420]
[Epoch 22/50] [Batch 937/938] [D loss: 1.324697, acc: 53%] [G loss: 1.663191]
[Epoch 23/50] [Batch 937/938] [D loss: 1.350422, acc: 60%] [G loss: 0.750597]
[Epoch 24/50] [Batch 937/938] [D loss: 1.246442, acc: 56%] [G loss: 0.938523]
[Epoch 25/50] [Batch 937/938] [D loss: 1.165384, acc: 57%] [G loss: 0.945472]
[Epoch 26/50] [Batch 937/938] [D loss: 1.318104, acc: 51%] [G loss: 0.943989]
[Epoch 27/50] [Batch 937/938] [D loss: 1.369688, acc: 51%] [G loss: 1.045065]
[Epoch 28/50] [Batch 937/938] [D loss: 1.241795, acc: 53%] [G loss: 0.753452]
[Epoch 29/50] [Batch 937/938] [D loss: 1.245152, acc: 59%] [G loss: 1.388610]
[Epoch 30/50] [Batch 937/938] [D loss: 1.415710, acc: 54%] [G loss: 1.065790]
[Epoch 31/50] [Batch 937/938] [D loss: 1.261829, acc: 56%] [G loss: 0.598533]
[Epoch 32/50] [Batch 937/938] [D loss: 1.087898, acc: 71%] [G loss: 1.146997]
[Epoch 33/50] [Batch 937/938] [D loss: 1.199838, acc: 53%] [G loss: 1.855803]
[Epoch 34/50] [Batch 937/938] [D loss: 1.146974, acc: 56%] [G loss: 0.522069]
[Epoch 35/50] [Batch 937/938] [D loss: 1.228530, acc: 57%] [G loss: 2.065096]
[Epoch 36/50] [Batch 937/938] [D loss: 1.220076, acc: 57%] [G loss: 1.654905]
[Epoch 37/50] [Batch 937/938] [D loss: 1.130150, acc: 62%] [G loss: 0.764476]
[Epoch 38/50] [Batch 937/938] [D loss: 1.200049, acc: 62%] [G loss: 0.843963]
[Epoch 39/50] [Batch 937/938] [D loss: 1.236222, acc: 62%] [G loss: 1.258294]
[Epoch 40/50] [Batch 937/938] [D loss: 1.133929, acc: 64%] [G loss: 0.937561]
[Epoch 41/50] [Batch 937/938] [D loss: 1.131473, acc: 73%] [G loss: 1.858397]
[Epoch 42/50] [Batch 937/938] [D loss: 1.208524, acc: 67%] [G loss: 2.926470]
[Epoch 43/50] [Batch 937/938] [D loss: 1.054991, acc: 70%] [G loss: 2.653653]
[Epoch 44/50] [Batch 937/938] [D loss: 1.598110, acc: 37%] [G loss: 0.824140]
[Epoch 45/50] [Batch 937/938] [D loss: 1.037547, acc: 76%] [G loss: 1.001997]
[Epoch 46/50] [Batch 937/938] [D loss: 1.346323, acc: 57%] [G loss: 0.364864]
[Epoch 47/50] [Batch 937/938] [D loss: 1.138918, acc: 71%] [G loss: 0.415268]
[Epoch 48/50] [Batch 937/938] [D loss: 0.977577, acc: 81%] [G loss: 2.182211]
[Epoch 49/50] [Batch 937/938] [D loss: 1.880296, acc: 42%] [G loss: 1.574159]

收获：未寻找到MNIST数据集，采用其他数据集进行测试，实验结果一般，后续会继续学习相关知识，寻找原因