第G7周：Semi-Supervised GAN 理论与实战

骑猪玩狗2024-09-09 12:42

SGAN概述

  • 简单来说，SGAN与普通GAN不同的点在于判别器
  • 普通的GAN的判别器只输出"真或假"
  • 而SGAN在输出真假的同时还充当分类器，将真分为真1、真2...

代码解析

配置代码

python 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

os.makedirs("images", exist_ok=True)

parser = argparse.ArgumentParser()
parser.add_argument("--n_epochs", type=int, default=50, help="number of epochs of training")
parser.add_argument("--batch_size", type=int, default=64, help="size of the batches")
parser.add_argument("--lr", type=float, default=0.0002, help="adam: learning rate")
parser.add_argument("--b1", type=float, default=0.5, help="adam: decay of first order momentum of gradient")
parser.add_argument("--b2", type=float, default=0.999, help="adam: decay of first order momentum of gradient")
parser.add_argument("--n_cpu", type=int, default=8, help="number of cpu threads to use during batch generation")
parser.add_argument("--latent_dim", type=int, default=100, help="dimensionality of the latent space")
parser.add_argument("--num_classes", type=int, default=10, help="number of classes for dataset")
parser.add_argument("--img_size", type=int, default=32, help="size of each image dimension")
parser.add_argument("--channels", type=int, default=1, help="number of image channels")
parser.add_argument("--sample_interval", type=int, default=400, help="interval between image sampling")
opt = parser.parse_args()
print(opt)

cuda = True if torch.cuda.is_available() else False

初始化权重

python 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)

定义模型

生成器

python 复制代码 
class Generator(nn.Module):
    def __init__(self):
        super(Generator, self).__init__()

        self.label_emb = nn.Embedding(opt.num_classes, opt.latent_dim)

        self.init_size = opt.img_size // 4  # Initial size before upsampling
        self.l1 = nn.Sequential(nn.Linear(opt.latent_dim, 128 * self.init_size ** 2))

        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

判别器

python 复制代码 
class Discriminator(nn.Module):
    def __init__(self):
        super(Discriminator, self).__init__()

        def discriminator_block(in_filters, out_filters, bn=True):
            """Returns layers of each discriminator block"""
            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

        self.conv_blocks = nn.Sequential(
            *discriminator_block(opt.channels, 16, bn=False),
            *discriminator_block(16, 32),
            *discriminator_block(32, 64),
            *discriminator_block(64, 128),
        )

        # The height and width of downsampled image
        ds_size = opt.img_size // 2 ** 4

        # Output layers
        self.adv_layer = nn.Sequential(nn.Linear(128 * ds_size ** 2, 1), nn.Sigmoid())
        self.aux_layer = nn.Sequential(nn.Linear(128 * ds_size ** 2, opt.num_classes + 1), nn.Softmax())

    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

配置模型

python 复制代码 
# Loss functions
adversarial_loss = torch.nn.BCELoss()
auxiliary_loss = torch.nn.CrossEntropyLoss()

# Initialize generator and discriminator
generator = Generator()
discriminator = Discriminator()

if cuda:
    generator.cuda()
    discriminator.cuda()
    adversarial_loss.cuda()
    auxiliary_loss.cuda()

# Initialize weights
generator.apply(weights_init_normal)
discriminator.apply(weights_init_normal)

# Configure data loader
os.makedirs("../../data/mnist", exist_ok=True)
dataloader = torch.utils.data.DataLoader(
    datasets.MNIST(
        "../../data/mnist",
        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,
)

# Optimizers
optimizer_G = torch.optim.Adam(generator.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))
optimizer_D = torch.optim.Adam(discriminator.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))

FloatTensor = torch.cuda.FloatTensor if cuda else torch.FloatTensor
LongTensor = torch.cuda.LongTensor if cuda else torch.LongTensor

训练模型

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

        batch_size = imgs.shape[0]

        # Adversarial ground truths
        valid = Variable(FloatTensor(batch_size, 1).fill_(1.0), requires_grad=False)
        fake = Variable(FloatTensor(batch_size, 1).fill_(0.0), requires_grad=False)
        fake_aux_gt = Variable(LongTensor(batch_size).fill_(opt.num_classes), requires_grad=False)

        # Configure input
        real_imgs = Variable(imgs.type(FloatTensor))
        labels = Variable(labels.type(LongTensor))

        # -----------------
        #  Train Generator
        # -----------------

        optimizer_G.zero_grad()

        # Sample noise and labels as generator input
        z = Variable(FloatTensor(np.random.normal(0, 1, (batch_size, opt.latent_dim))))

        # Generate a batch of images
        gen_imgs = generator(z)

        # Loss measures generator's ability to fool the discriminator
        validity, _ = discriminator(gen_imgs)
        g_loss = adversarial_loss(validity, valid)

        g_loss.backward()
        optimizer_G.step()

        # ---------------------
        #  Train Discriminator
        # ---------------------

        optimizer_D.zero_grad()

        # Loss for real images
        real_pred, real_aux = discriminator(real_imgs)
        d_real_loss = (adversarial_loss(real_pred, valid) + auxiliary_loss(real_aux, labels)) / 2

        # Loss for fake images
        fake_pred, fake_aux = discriminator(gen_imgs.detach())
        d_fake_loss = (adversarial_loss(fake_pred, fake) + auxiliary_loss(fake_aux, fake_aux_gt)) / 2

        # Total discriminator loss
        d_loss = (d_real_loss + d_fake_loss) / 2

        # Calculate discriminator accuracy
        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())
    )
运行过程
python 复制代码 
PS F:\365data\G7> & E:/anaconda3/envs/PGPU/python.exe f:/365data/G7/sgan.py
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)
E:\anaconda3\envs\PGPU\lib\site-packages\torch\nn\modules\container.py:139: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.
  input = module(input)
[Epoch 0/50] [Batch 937/938] [D loss: 1.359867, acc: 50%] [G loss: 0.656197]
[Epoch 1/50] [Batch 937/938] [D loss: 1.358995, acc: 50%] [G loss: 0.720781]
[Epoch 2/50] [Batch 937/938] [D loss: 1.344833, acc: 48%] [G loss: 0.804772]
[Epoch 3/50] [Batch 937/938] [D loss: 1.333392, acc: 54%] [G loss: 0.688524]
[Epoch 4/50] [Batch 937/938] [D loss: 1.382262, acc: 50%] [G loss: 0.701481]
[Epoch 5/50] [Batch 937/938] [D loss: 1.269132, acc: 62%] [G loss: 0.913967]
[Epoch 6/50] [Batch 937/938] [D loss: 1.314945, acc: 50%] [G loss: 0.828390]
[Epoch 7/50] [Batch 937/938] [D loss: 1.410788, acc: 50%] [G loss: 0.852402]
[Epoch 8/50] [Batch 937/938] [D loss: 1.447273, acc: 46%] [G loss: 1.019709]
[Epoch 9/50] [Batch 937/938] [D loss: 1.314910, acc: 54%] [G loss: 0.726437]
[Epoch 10/50] [Batch 937/938] [D loss: 1.404197, acc: 48%] [G loss: 0.811808]
[Epoch 11/50] [Batch 937/938] [D loss: 1.260281, acc: 59%] [G loss: 0.698684]
[Epoch 12/50] [Batch 937/938] [D loss: 1.423180, acc: 51%] [G loss: 0.661626]
[Epoch 13/50] [Batch 937/938] [D loss: 1.276824, acc: 50%] [G loss: 1.359684]
[Epoch 14/50] [Batch 937/938] [D loss: 1.351237, acc: 54%] [G loss: 0.595949]
[Epoch 15/50] [Batch 937/938] [D loss: 1.398958, acc: 53%] [G loss: 0.947947]
[Epoch 16/50] [Batch 937/938] [D loss: 1.282681, acc: 50%] [G loss: 1.282415]
[Epoch 17/50] [Batch 937/938] [D loss: 1.248466, acc: 56%] [G loss: 0.875078]
[Epoch 18/50] [Batch 937/938] [D loss: 1.220254, acc: 54%] [G loss: 0.849395]
[Epoch 19/50] [Batch 937/938] [D loss: 1.298958, acc: 51%] [G loss: 1.156524]
[Epoch 20/50] [Batch 937/938] [D loss: 1.243778, acc: 57%] [G loss: 0.959405]
[Epoch 21/50] [Batch 937/938] [D loss: 1.497701, acc: 50%] [G loss: 1.469938]
[Epoch 22/50] [Batch 937/938] [D loss: 1.303683, acc: 51%] [G loss: 1.698360]
[Epoch 23/50] [Batch 937/938] [D loss: 1.312811, acc: 48%] [G loss: 0.858052]
[Epoch 24/50] [Batch 937/938] [D loss: 1.176002, acc: 56%] [G loss: 1.668655]
[Epoch 25/50] [Batch 937/938] [D loss: 1.246156, acc: 62%] [G loss: 0.799746]
[Epoch 26/50] [Batch 937/938] [D loss: 1.317814, acc: 56%] [G loss: 2.129106]
[Epoch 27/50] [Batch 937/938] [D loss: 1.405645, acc: 45%] [G loss: 0.796651]
[Epoch 28/50] [Batch 937/938] [D loss: 1.295846, acc: 56%] [G loss: 1.046427]
[Epoch 29/50] [Batch 937/938] [D loss: 1.188294, acc: 57%] [G loss: 1.462691]
[Epoch 30/50] [Batch 937/938] [D loss: 1.141519, acc: 65%] [G loss: 1.202172]
[Epoch 31/50] [Batch 937/938] [D loss: 1.252719, acc: 60%] [G loss: 0.761035]
[Epoch 32/50] [Batch 937/938] [D loss: 1.182667, acc: 62%] [G loss: 1.138950]
[Epoch 33/50] [Batch 937/938] [D loss: 1.138652, acc: 57%] [G loss: 1.878641]
[Epoch 34/50] [Batch 937/938] [D loss: 1.614118, acc: 40%] [G loss: 1.174179]
[Epoch 35/50] [Batch 937/938] [D loss: 1.127694, acc: 59%] [G loss: 1.391736]
[Epoch 36/50] [Batch 937/938] [D loss: 1.144518, acc: 67%] [G loss: 0.531392]
[Epoch 37/50] [Batch 937/938] [D loss: 1.133765, acc: 59%] [G loss: 0.514877]
[Epoch 38/50] [Batch 937/938] [D loss: 1.562284, acc: 48%] [G loss: 2.186804]
[Epoch 39/50] [Batch 937/938] [D loss: 1.197355, acc: 67%] [G loss: 1.720435]
[Epoch 40/50] [Batch 937/938] [D loss: 1.140882, acc: 62%] [G loss: 2.072633]
[Epoch 41/50] [Batch 937/938] [D loss: 1.180019, acc: 56%] [G loss: 2.425399]
[Epoch 42/50] [Batch 937/938] [D loss: 1.028821, acc: 73%] [G loss: 3.478460]
[Epoch 43/50] [Batch 937/938] [D loss: 1.099987, acc: 76%] [G loss: 0.333971]
[Epoch 44/50] [Batch 937/938] [D loss: 1.421913, acc: 51%] [G loss: 0.661157]
[Epoch 45/50] [Batch 937/938] [D loss: 1.300432, acc: 54%] [G loss: 0.675104]
[Epoch 46/50] [Batch 937/938] [D loss: 1.011263, acc: 70%] [G loss: 0.876295]
[Epoch 47/50] [Batch 937/938] [D loss: 1.159203, acc: 70%] [G loss: 1.982926]
[Epoch 48/50] [Batch 937/938] [D loss: 1.027292, acc: 71%] [G loss: 0.962503]
[Epoch 49/50] [Batch 937/938] [D loss: 0.890597, acc: 87%] [G loss: 0.349576]
运行结果



总结

  • SGAN在判别器的输出中增加了一个分支，用以对"真"进行分类
python 复制代码 
        # 判断真假
        self.adv_layer = nn.Sequential(nn.Linear(128 * ds_size ** 2, 1), nn.Sigmoid())
        # 判断类别
        self.aux_layer = nn.Sequential(nn.Linear(128 * ds_size ** 2, opt.num_classes + 1), nn.Softmax())
  • 由于aux_layer是多分类输出，因此激活函数使用Softmax，其输出是个概率分布，总和为1
相关推荐
love you joyfully3 分钟前
目标检测与R-CNN——paddle部分
人工智能·目标检测·cnn·paddle
AI视觉网奇23 分钟前
Detected at node ‘truediv‘ defined at (most recent call last): Node: ‘truediv‘
人工智能·python·tensorflow
西西弗Sisyphus26 分钟前
开放世界目标检测 Grounding DINO
人工智能·目标检测·计算机视觉·大模型
抓哇能手44 分钟前
数据库系统概论
数据库·人工智能·sql·mysql·计算机
火云洞红孩儿1 小时前
基于AI IDE 打造快速化的游戏LUA脚本的生成系统
c++·人工智能·inscode·游戏引擎·lua·游戏开发·脚本系统
风清扬雨1 小时前
【计算机视觉】超简单！傅里叶变换的经典案例
人工智能·计算机视觉
HuggingFace1 小时前
自动评估基准 | 设计你的自动评估任务
人工智能·自动评估
GISer_Jing2 小时前
神经网络初学总结（一）
人工智能·深度学习·神经网络
szxinmai主板定制专家2 小时前
【国产NI替代】基于A7 FPGA+AI的16振动（16bits）终端PCIE数据采集板卡
人工智能·fpga开发
热门推荐
0119个Web前端交互式3D JavaScript框架和库02ARM学习（31）编译器对overlay方式的支持03玄机平台应急响应—webshell查杀04RAG 实践- Ollama+RagFlow 部署本地知识库05【一文读懂】NTN（非地面网络）技术介绍06(欧拉)openEuler系统添加网卡文件配置流程、(欧拉)openEuler系统手动配置ipv6地址流程、(欧拉)openEuler系统网络管理说明07Ubuntu 20.04使用Livox mid 360 测试 FAST_LIO08组基轨迹建模 GBTM的介绍与实现（Stata 或 R）09Docker 夺命连环 15 问10大模型国产化适配7-华为昇腾LLM落地可选解决方案（MindFormers、ModelLink、MindIE）