LSTM-GAN读取11*11灰度图像文件夹并将生成的每个不同图像保存 代码实现

以下是使用LSTM-GAN生成11*11灰度图像并保存的代码实现，注释中有详细的解释：

import numpy as np
from tensorflow.keras.layers import Input, Dense, Reshape, LSTM, Dropout, LeakyReLU, UpSampling2D, Conv2D, Flatten
from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.optimizers import Adam
import os
import cv2

# 设置图像大小和通道数
img_rows = 11
img_cols = 11
channels = 1

# 定义生成器模型
def build_generator():

    model = Sequential()

    # 添加LSTM层
    model.add(LSTM(128, input_shape=(100,)))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Dense(11*11*1, activation='tanh'))
    model.add(Reshape((11, 11, 1)))

    return model

# 定义判别器模型
def build_discriminator():

    model = Sequential()

    # 添加卷积层
    model.add(Conv2D(32, kernel_size=3, strides=2, input_shape=(11, 11, 1), padding="same"))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Dropout(0.25))
    model.add(Conv2D(64, kernel_size=3, strides=2, padding="same"))
    model.add(ZeroPadding2D(padding=((0,1),(0,1))))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Dropout(0.25))
    model.add(Conv2D(128, kernel_size=3, strides=2, padding="same"))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Dropout(0.25))
    model.add(Conv2D(256, kernel_size=3, strides=1, padding="same"))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Dropout(0.25))
    model.add(Flatten())
    model.add(Dense(1, activation='sigmoid'))

    return model

# 定义LSTM-GAN模型
def build_lstm_gan(generator, discriminator):

    # 冻结判别器模型
    discriminator.trainable = False

    # 定义输入和噪声
    gan_input = Input(shape=(100,))
    x = generator(gan_input)
    gan_output = discriminator(x)

    # 定义LSTM-GAN模型
    gan = Model(inputs=gan_input, outputs=gan_output)

    return gan

# 加载数据集
def load_data():
    
    data = []
    
    # 读取灰度图像文件夹
    for filename in os.listdir('gray_images'):
        img = cv2.imread(os.path.join('gray_images', filename), cv2.IMREAD_GRAYSCALE)
        img = cv2.resize(img, (img_rows, img_cols))
        img = img.astype('float32') / 127.5 - 1.
        img = np.expand_dims(img, axis=2)
        data.append(img)

    return np.array(data)

# 训练LSTM-GAN模型
def train():

    # 加载数据集
    X_train = load_data()

    # 编译判别器模型
    discriminator = build_discriminator()
    discriminator.compile(loss='binary_crossentropy', optimizer=Adam(), metrics=['accuracy'])

    # 编译生成器模型
    generator = build_generator()

    # 编译LSTM-GAN模型
    lstm_gan = build_lstm_gan(generator, discriminator)
    lstm_gan.compile(loss='binary_crossentropy', optimizer=Adam())

    # 定义噪声维度
    noise_dim = 100

    # 定义训练参数
    epochs = 100
    batch_size = 128
    save_interval = 10

    # 定义标签
    real_labels = np.ones((batch_size, 1))
    fake_labels = np.zeros((batch_size, 1))

    for epoch in range(epochs):

        # 随机选择一批样本
        idx = np.random.randint(0, X_train.shape[0], batch_size)
        real_imgs = X_train[idx]

        # 生成噪声
        noise = np.random.normal(0, 1, (batch_size, noise_dim))

        # 生成假图像
        fake_imgs = generator.predict(noise)

        # 训练判别器
        d_loss_real = discriminator.train_on_batch(real_imgs, real_labels)
        d_loss_fake = discriminator.train_on_batch(fake_imgs, fake_labels)
        d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)

        # 训练生成器
        g_loss = lstm_gan.train_on_batch(noise, real_labels)

        # 打印训练进度
        print("Epoch %d [D loss: %f, acc.: %.2f%%] [G loss: %f]" % (epoch, d_loss[0], 100*d_loss[1], g_loss))

        # 保存生成器模型
        if epoch % save_interval == 0:
            generator.save('generator_%d.h5' % epoch)

            # 生成并保存图像
            r, c = 5, 5
            noise = np.random.normal(0, 1, (r * c, noise_dim))
            gen_imgs = generator.predict(noise)
            gen_imgs = 0.5 * gen_imgs + 0.5
            for i in range(r * c):
                cv2.imwrite("images/%d.png" % i, gen_imgs[i,:,:,0] * 255)

代码实现了以下功能：

1. 定义了生成器模型、判别器模型和LSTM-GAN模型。
2. 加载11*11的灰度图像文件夹，并将所有图像统一缩放为11*11大小的灰度图像。
3. 训练LSTM-GAN模型并保存生成器模型，每10个epoch保存一次，并生成一些生成的图像并保存在images文件夹中。