第G6周：CycleGan实战

最新推荐文章于 2024-04-13 09:44:07 发布

千筱夜

最新推荐文章于 2024-04-13 09:44:07 发布

阅读量1k

点赞数 26

文章标签：深度学习

本文链接：https://blog.csdn.net/geo436872/article/details/137674819

版权

🍨 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

Cyclegan应用有哪些？

图像风格转换：使用CycleGAN可以将一种图像风格转换成另一种图像风格，例如将普通照片转换成名画的风格。

图像颜色化：使用CycleGAN可以将黑白照片转换成彩色照片。

季节转换：使用CycleGAN可以将夏天的风景转换成冬天的风景，或者将秋天的风景转换成春天的风景。

域适应：使用CycleGAN可以将一个域的数据转换成另一个域的数据，以便在目标域上进行更好的训练和测试。

超分辨率：使用CycleGAN可以将低分辨率的图像转换成高分辨率的图像。

医学图像处理：使用CycleGAN可以将医学图像进行增强和改善，以便于医生更好地进行诊断和治疗。

CycleGan与Gan差异有哪些？

目标不同：GAN的目标是生成与真实数据分布相似的数据，而CycleGAN的目标是将一个域的数据映射到另一个域的数据分布。

训练方式不同：GAN使用单个生成器和单个判别器进行训练，而CycleGAN使用两个生成器和两个判别器进行训练，并且需要满足循环一致性约束。

应用场景不同：GAN通常用于生成新的数据样本，例如人脸、音乐等，而CycleGAN通常用于图像风格转换、图像颜色化、季节转换等任务。

结构不同：GAN通常由一个生成器和一个判别器组成，生成器负责生成数据，判别器负责区分真实数据和生成数据。而CycleGAN由两个生成器和两个判别器组成，每个生成器负责将一个域的数据映射到另一个域，每个判别器负责区分来自不同域的数据。

损失函数不同：GAN的损失函数包括生成器损失和判别器损失，而CycleGAN的损失函数包括生成器损失、循环一致性损失和判别器损失。

CycleGan具体架构简介

在这里插入图片描述

CycleGAN的核心思想是通过两个生成器和两个判别器的组合，实现图像的风格转换，即将一种领域的图像转换为另一种领域的图像，同时保持内容不变。其中，CycleGAN的生成器和判别器分别如下：

生成器：包括一个从A到B的生成器 $G_{A \rightarrow B}$ 和一个从B到A的生成器 $G_{B \rightarrow A}$ 。这两个生成器分别负责将A领域的图像转换为B领域的图像，以及将B领域的图像转换为A领域的图像。
判别器：包括一个针对A领域的判别器 $D_A$ 和一个针对B领域的判别器 $D_B$ 。这两个判别器分别负责判断输入的图像是来自A领域还是B领域。

CycleGAN的损失函数主要包括以下几部分：

对抗损失(Adversarial Loss)：生成器和判别器之间的对抗损失，帮助生成器生成更逼真的图像，同时训练判别器来区分真实和生成的图像。
循环一致性损失(Cycle-Consistency Loss)：通过循环一致性损失来保证转换的图像能够保持原始图像的内容，即将A到B再到A的转换结果与原始输入A之间的差异最小化。
同一性损失(Identity Loss)：确保输入图像经过生成器转换后与自身的差异最小化，避免图像内容的丢失。

训练

在这里插入图片描述
data文件夹里面放数据集，images里面放保存的图片，save_models放保存的模型。

cyclegan.py

import argparse
import itertools
from torchvision.utils import save_image, make_grid
from torch.utils.data import DataLoader
from models import *
from datasets import *
from utils import *
import torch
import PIL
from PIL import Image


parser = argparse.ArgumentParser()
parser.add_argument("--epoch", type=int, default=0, help="epoch to start training from")
parser.add_argument("--n_epochs", type=int, default=2, help="number of epochs of training")
parser.add_argument("--dataset_name", type=str, default="monet2photo", help="name of the dataset")
parser.add_argument("--batch_size", type=int, default=1, 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("--decay_epoch", type=int, default=1, help="epoch from which to start lr decay")
parser.add_argument("--n_cpu", type=int, default=4, help="number of cpu threads to use during batch generation")
parser.add_argument("--img_height", type=int, default=256, help="size of image height")
parser.add_argument("--img_width", type=int, default=256, help="size of image width")
parser.add_argument("--channels", type=int, default=3, help="number of image channels")
parser.add_argument("--sample_interval", type=int, default=100, help="interval between saving generator outputs")
parser.add_argument("--checkpoint_interval", type=int, default=1, help="interval between saving model checkpoints")
parser.add_argument("--n_residual_blocks", type=int, default=9, help="number of residual blocks in generator")
parser.add_argument("--lambda_cyc", type=float, default=10.0, help="cycle loss weight")
parser.add_argument("--lambda_id", type=float, default=5.0, help="identity loss weight")
opt = parser.parse_args()
print(opt)

# Create sample and checkpoint directories
os.makedirs("images/%s" % opt.dataset_name, exist_ok=True)
os.makedirs("saved_models/%s" % opt.dataset_name, exist_ok=True)

# Losses
criterion_GAN = torch.nn.MSELoss()
criterion_cycle = torch.nn.L1Loss()
criterion_identity = torch.nn.L1Loss()

cuda = torch.cuda.is_available()
input_shape = (opt.channels, opt.img_height, opt.img_width)

# 初始化生成器鉴别器
G_AB = GeneratorResNet(input_shape, opt.n_residual_blocks)
G_BA = GeneratorResNet(input_shape, opt.n_residual_blocks)
D_A = Discriminator(input_shape)
D_B = Discriminator(input_shape)

if cuda:
    G_AB = G_AB.cuda()
    G_BA = G_BA.cuda()
    D_A = D_A.cuda()
    D_B = D_B.cuda()
    criterion_GAN.cuda()
    criterion_cycle.cuda()
    criterion_identity.cuda()

if opt.epoch != 0:
    # 加载预训练模型
    G_AB.load_state_dict(torch.load("saved_models/%s/G_AB_%d.pth" % (opt.dataset_name, opt.epoch)))
    G_BA.load_state_dict(torch.load("saved_models/%s/G_BA_%d.pth" % (opt.dataset_name, opt.epoch)))
    D_A.load_state_dict(torch.load("saved_models/%s/D_A_%d.pth" % (opt.dataset_name, opt.epoch)))
    D_B.load_state_dict(torch.load("saved_models/%s/D_B_%d.pth" % (opt.dataset_name, opt.epoch)))
else:
    # 初始化权重
    G_AB.apply(weights_init_normal)
    G_BA.apply(weights_init_normal)
    D_A.apply(weights_init_normal)
    D_B.apply(weights_init_normal)

# Optimizers
optimizer_G = torch.optim.Adam(
    itertools.chain(G_AB.parameters(), G_BA.parameters()), lr=opt.lr, betas=(opt.b1, opt.b2)
)
optimizer_D_A = torch.optim.Adam(D_A.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))
optimizer_D_B = torch.optim.Adam(D_B.parameters(), lr=opt.lr, betas=(opt.b1, opt.b2))

# Learning rate update schedulers
lr_scheduler_G = torch.optim.lr_scheduler.LambdaLR(
    optimizer_G, lr_lambda=LambdaLR(opt.n_epochs, opt.epoch, opt.decay_epoch).step
)
lr_scheduler_D_A = torch.optim.lr_scheduler.LambdaLR(
    optimizer_D_A, lr_lambda=LambdaLR(opt.n_epochs, opt.epoch, opt.decay_epoch).step
)
lr_scheduler_D_B = torch.optim.lr_scheduler.LambdaLR(
    optimizer_D_B, lr_lambda=LambdaLR(opt.n_epochs, opt.epoch, opt.decay_epoch).step
)

Tensor = torch.cuda.FloatTensor if cuda else torch.Tensor

# Buffers of previously generated samples
fake_A_buffer = ReplayBuffer()
fake_B_buffer = ReplayBuffer()

# Image transformations
transforms_ = [
    transforms.Resize(int(opt.img_height * 1.12), PIL.Image.BICUBIC),
    transforms.RandomCrop((opt.img_height, opt.img_width)),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
]

# Training data loader
dataloader = DataLoader(
    ImageDataset("./data/%s/" % opt.dataset_name, transforms_=transforms_, unaligned=True),
    batch_size=opt.batch_size,
    shuffle=True,
    num_workers=opt.n_cpu,
)
# Test data loader
val_dataloader = DataLoader(
    ImageDataset("./data/%s/" % opt.dataset_name, transforms_=transforms_, unaligned=True, mode="test"),
    batch_size=5,
    shuffle=True,
    num_workers=1,
)


def sample_images(batches_done):
    """Saves a generated sample from the test set"""
    imgs = next(iter(val_dataloader))
    G_AB.eval()
    G_BA.eval()
    real_A = Variable(imgs["A"].type(Tensor))
    fake_B = G_AB(real_A)
    real_B = Variable(imgs["B"].type(Tensor))
    fake_A = G_BA(real_B)
    # Arange images along x-axis
    real_A = make_grid(real_A, nrow=5, normalize=True)
    real_B = make_grid(real_B, nrow=5, normalize=True)
    fake_A = make_grid(fake_A, nrow=5, normalize=True)
    fake_B = make_grid(fake_B, nrow=5, normalize=True)
    # Arange images along y-axis
    image_grid = torch.cat((real_A, fake_B, real_B, fake_A), 1)
    save_image(image_grid, "images/%s/%s.png" % (opt.dataset_name, batches_done), normalize=False)


# ----------
#  Training
# ----------


if __name__ == '__main__':

    prev_time = time.time()
    for epoch in range(opt.epoch, opt.n_epochs):
        for i, batch in enumerate(dataloader):

            # Set model input
            real_A = Variable(batch["A"].type(Tensor))
            real_B = Variable(batch["B"].type(Tensor))

            # Adversarial ground truths
            valid = Variable(Tensor(np.ones((real_A.size(0), *D_A.output_shape))), requires_grad=False)
            fake  = Variable(Tensor(np.zeros((real_A.size(0), *D_A.output_shape))), requires_grad=False)

            # ------------------
            #  Train Generators
            # ------------------

            G_AB.train()
            G_BA.train()

            optimizer_G.zero_grad()

            # Identity loss
            loss_id_A = criterion_identity(G_BA(real_A), real_A)
            loss_id_B = criterion_identity(G_AB(real_B), real_B)

            loss_identity = (loss_id_A + loss_id_B) / 2

            # GAN loss
            fake_B = G_AB(real_A)
            loss_GAN_AB = criterion_GAN(D_B(fake_B), valid)
            fake_A = G_BA(real_B)
            loss_GAN_BA = criterion_GAN(D_A(fake_A), valid)

            loss_GAN = (loss_GAN_AB + loss_GAN_BA) / 2

            # Cycle loss
            recov_A = G_BA(fake_B)
            loss_cycle_A = criterion_cycle(recov_A, real_A)
            recov_B = G_AB(fake_A)
            loss_cycle_B = criterion_cycle(recov_B, real_B)

            loss_cycle = (loss_cycle_A + loss_cycle_B) / 2

            # Total loss
            loss_G = loss_GAN + opt.lambda_cyc * loss_cycle + opt.lambda_id * loss_identity

            loss_G.backward()
            optimizer_G.step()

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

            optimizer_D_A.zero_grad()

            # Real loss
            loss_real = criterion_GAN(D_A(real_A), valid)
            # Fake loss (on batch of previously generated samples)
            fake_A_ = fake_A_buffer.push_and_pop(fake_A)
            loss_fake = criterion_GAN(D_A(fake_A_.detach()), fake)
            # Total loss
            loss_D_A = (loss_real + loss_fake) / 2

            loss_D_A.backward()
            optimizer_D_A.step()

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

            optimizer_D_B.zero_grad()

            # Real loss
            loss_real = criterion_GAN(D_B(real_B), valid)
            # Fake loss (on batch of previously generated samples)
            fake_B_ = fake_B_buffer.push_and_pop(fake_B)
            loss_fake = criterion_GAN(D_B(fake_B_.detach()), fake)
            # Total loss
            loss_D_B = (loss_real + loss_fake) / 2

            loss_D_B.backward()
            optimizer_D_B.step()

            loss_D = (loss_D_A + loss_D_B) / 2

            # --------------
            #  Log Progress
            # --------------

            # Determine approximate time left
            batches_done = epoch * len(dataloader) + i
            batches_left = opt.n_epochs * len(dataloader) - batches_done
            time_left = datetime.timedelta(seconds=batches_left * (time.time() - prev_time))
            prev_time = time.time()

            # Print log
            sys.stdout.write(
                "\r[Epoch %d/%d] [Batch %d/%d] [D loss: %f] [G loss: %f, adv: %f, cycle: %f, identity: %f] ETA: %s"
                % (
                    epoch,
                    opt.n_epochs,
                    i,
                    len(dataloader),
                    loss_D.item(),
                    loss_G.item(),
                    loss_GAN.item(),
                    loss_cycle.item(),
                    loss_identity.item(),
                    time_left,
                )
            )

            # If at sample interval save image
            if batches_done % opt.sample_interval == 0:
                sample_images(batches_done)

        # Update learning rates
        lr_scheduler_G.step()
        lr_scheduler_D_A.step()
        lr_scheduler_D_B.step()

        if opt.checkpoint_interval != -1 and epoch % opt.checkpoint_interval == 0:
            # Save model checkpoints
            torch.save(G_AB.state_dict(), "saved_models/%s/G_AB_%d.pth" % (opt.dataset_name, epoch))
            torch.save(G_BA.state_dict(), "saved_models/%s/G_BA_%d.pth" % (opt.dataset_name, epoch))
            torch.save(D_A.state_dict(), "saved_models/%s/D_A_%d.pth" % (opt.dataset_name, epoch))
            torch.save(D_B.state_dict(), "saved_models/%s/D_B_%d.pth" % (opt.dataset_name, epoch))

model

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


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)
        if hasattr(m, "bias") and m.bias is not None:
            torch.nn.init.constant_(m.bias.data, 0.0)
    elif classname.find("BatchNorm2d") != -1:
        torch.nn.init.normal_(m.weight.data, 1.0, 0.02)
        torch.nn.init.constant_(m.bias.data, 0.0)


##############################
#           RESNET
##############################


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

        self.block = nn.Sequential(
            nn.ReflectionPad2d(1),
            nn.Conv2d(in_features, in_features, 3),
            nn.InstanceNorm2d(in_features),
            nn.ReLU(inplace=True),
            nn.ReflectionPad2d(1),
            nn.Conv2d(in_features, in_features, 3),
            nn.InstanceNorm2d(in_features),
        )

    def forward(self, x):
        return x + self.block(x)


class GeneratorResNet(nn.Module):
    def __init__(self, input_shape, num_residual_blocks):
        super(GeneratorResNet, self).__init__()

        channels = input_shape[0]

        # Initial convolution block
        out_features = 64
        model = [
            nn.ReflectionPad2d(channels),
            nn.Conv2d(channels, out_features, 7),
            nn.InstanceNorm2d(out_features),
            nn.ReLU(inplace=True),
        ]
        in_features = out_features

        # Downsampling
        for _ in range(2):
            out_features *= 2
            model += [
                nn.Conv2d(in_features, out_features, 3, stride=2, padding=1),
                nn.InstanceNorm2d(out_features),
                nn.ReLU(inplace=True),
            ]
            in_features = out_features

        # Residual blocks
        for _ in range(num_residual_blocks):
            model += [ResidualBlock(out_features)]

        # Upsampling
        for _ in range(2):
            out_features //= 2
            model += [
                nn.Upsample(scale_factor=2),
                nn.Conv2d(in_features, out_features, 3, stride=1, padding=1),
                nn.InstanceNorm2d(out_features),
                nn.ReLU(inplace=True),
            ]
            in_features = out_features

        # Output layer
        model += [nn.ReflectionPad2d(channels), nn.Conv2d(out_features, channels, 7), nn.Tanh()]

        self.model = nn.Sequential(*model)

    def forward(self, x):
        return self.model(x)


##############################
#        Discriminator
##############################


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

        channels, height, width = input_shape

        # Calculate output shape of image discriminator (PatchGAN)
        self.output_shape = (1, height // 2 ** 4, width // 2 ** 4)

        def discriminator_block(in_filters, out_filters, normalize=True):
            """Returns downsampling layers of each discriminator block"""
            layers = [nn.Conv2d(in_filters, out_filters, 4, stride=2, padding=1)]
            if normalize:
                layers.append(nn.InstanceNorm2d(out_filters))
            layers.append(nn.LeakyReLU(0.2, inplace=True))
            return layers

        self.model = nn.Sequential(
            *discriminator_block(channels, 64, normalize=False),
            *discriminator_block(64, 128),
            *discriminator_block(128, 256),
            *discriminator_block(256, 512),
            nn.ZeroPad2d((1, 0, 1, 0)),
            nn.Conv2d(512, 1, 4, padding=1)
        )

    def forward(self, img):
        return self.model(img)

最开始因为虚拟内存不够无法正常训练，于是将n_cpu调小至4，可以运行，运行结果等出来后补上。在这里插入图片描述

千筱夜

关注

26
点赞
踩
20

收藏

觉得还不错? 一键收藏
0
评论
第G6周：CycleGan实战

CycleGAN的核心思想是通过两个生成器和两个判别器的组合，实现图像的风格转换，即将一种领域的图像转换为另一种领域的图像，同时保持内容不变。其中，CycleGAN的生成器和判别器分别如下：生成器：包括一个从A到B的生成器GA→BGA→B和一个从B到A的生成器GB→AGB→A。这两个生成器分别负责将A领域的图像转换为B领域的图像，以及将B领域的图像转换为A领域的图像。判别器：包括一个针对A领域的判别器DAD_ADA和一个针对B领域的判别器DB。
复制链接

扫一扫