我在训练我的 GAN 模型时一次又一次地面临 ValueError 的问题。请帮助我。
import cv2
import numpy as np
import pandas as pd
import tensorflow as tf
# read the CSV file containing the labels
labels_df = pd.read_csv('labels.csv')
# define a function to preprocess and resize the images
def preprocess_image(image, target_size):
# convert the image to grayscale
gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# resize the image to the target size
resized_image = cv2.resize(gray_image, target_size)
# normalize the pixel values to be between 0 and 1
normalized_image = resized_image / 255.0
return normalized_image
# initialize lists to store the images and labels
images = []
labels = []
# define the target size of the resized images
target_size = (128, 128)
# iterate over the rows of the dataframe and read the corresponding image using OpenCV
for index, row in labels_df.iterrows():
image_filename = row['image_filename']
label = row['label']
# read image using OpenCV
image = cv2.imread(image_filename)
# preprocess and resize the image
preprocessed_image = preprocess_image(image, target_size)
# save the preprocessed image and label in numpy arrays
images.append(preprocessed_image)
labels.append(label)
# convert the lists to numpy arrays
images = np.array(images)
labels = np.array(labels)
# save the numpy arrays to disk
np.save('preprocessed_images.npy', images)
np.save('labels.npy', labels)
# load the preprocessed images and labels from the numpy arrays
images = np.load('preprocessed_images.npy')
labels = np.load('labels.npy')
# define the input shape of the generator
generator_input_shape = 100
# define the generator model
generator = tf.keras.Sequential([
tf.keras.layers.Dense(7 * 7 * 64, input_dim=generator_input_shape),
tf.keras.layers.Reshape((7, 7, 64)),
tf.keras.layers.Conv2DTranspose(64, kernel_size=4, strides=2, padding='same'),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(alpha=0.2),
tf.keras.layers.Conv2DTranspose(64, kernel_size=4, strides=2, padding='same'),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(alpha=0.2),
tf.keras.layers.Conv2D(1, kernel_size=7, activation='sigmoid', padding='same')
])
# define the discriminator model
discriminator = tf.keras.Sequential([
tf.keras.layers.Conv2D(64, kernel_size=4, strides=2, input_shape=(28, 28, 1), padding='same'),
tf.keras.layers.LeakyReLU(alpha=0.2),
tf.keras.layers.Dropout(0.4),
tf.keras.layers.Conv2D(128, kernel_size=4, strides=2, padding='same'),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.LeakyReLU(alpha=0.2),
tf.keras.layers.Dropout(0.4),
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(1, activation='sigmoid')
])
# define the batch size and number of epochs
batch_size = 12
epochs = 10
# define the labels for the real and fake images
real_labels = np.ones((batch_size, 1))
fake_labels = np.zeros((batch_size, 1))
# define the GAN model by combining the generator and discriminator
gan = tf.keras.Sequential([generator, discriminator])
# compile the GAN model
discriminator.compile(loss='binary_crossentropy', optimizer=tf.keras.optimizers.Adam(lr=0.0002, beta_1=0.5))
gan.compile(loss='binary_crossentropy', optimizer=tf.keras.optimizers.Adam(lr=0.0002, beta_1=0.5))
# iterate over the epochs
for epoch in range(epochs):
# select a random batch of images
idx = np.random.randint(0, images.shape[0], batch_size)
real_images = images[idx].reshape(batch_size, 28, 28, 1) # corrected
# generate a batch of fake images
noise = np.random.normal(0, 1, (batch_size, generator_input_shape))
fake_images = generator.predict(noise)
# train the discriminator on real and fake images
real_labels = np.ones((batch_size, 1))
fake_labels = np.zeros((batch_size, 1))
d_loss_real = discriminator.train_on_batch(real_images, real_labels)
d_loss_fake = discriminator.train_on_batch(fake_images, fake_labels)
# calculate the discriminator loss for this batch
d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)
# train the generator to fool the discriminator
noise = np.random.normal(0, 1, (batch_size, generator_input_shape))
fool_labels = np.ones((batch_size, 1))
g_loss = gan.train_on_batch(noise, fool_labels)
# print the progress
print(f"Epoch {epoch+1}/{epochs} | Discriminator loss: {d_loss[0]} | Generator loss: {g_loss}")
discriminator.summary()
generator.summary()
这是代码。
错误信息： ”------------------------------------------------ -------------------------- ValueError Traceback（最后一次调用） 单元格输入 [27]，第 53 行 范围内的纪元为 50（纪元）： 51 # 选择一批随机图像 52 idx = np.random.randint(0, 图像.shape[0], batch_size) ---> 53 real_images = images[idx].reshape(batch_size, 28, 28, 1) # 更正 55 # 生成一批假图像 56 噪声 = np.random.normal(0, 1, (batch_size, generator_input_shape))
生成器总结： “ 型号：“sequential_9”
dense_6（密集）（无，3136）316736
reshape_3（重塑）（无、7、7、64）0
conv2d_transpose_6 (Conv2DT (None, 14, 14, 64) 65600
转置）
batch_normalization_9 (Batc (None, 14, 14, 64) 256
h归一化）
leaky_re_lu_12 (LeakyReLU) (无, 14, 14, 64) 0
conv2d_transpose_7 (Conv2DT (None, 28, 28, 64) 65600
转置）
batch_normalization_10（蝙蝠（无、28、28、64）256
ch归一化）
leaky_re_lu_13（LeakyReLU）（无、28、28、64）0
conv2d_9 (Conv2D) (无, 28, 28, 1) 3137
================================================= ================ 总参数：451,585 可训练参数：451,329 不可训练的参数：256 ______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
鉴别器总结：
“型号：”sequential_10“
conv2d_10 (Conv2D) (无, 14, 14, 64) 1088
leaky_re_lu_14 (LeakyReLU) (无, 14, 14, 64) 0
dropout_6（辍学）（无，14、14、64）0
conv2d_11 (Conv2D) (无, 7, 7, 128) 131200
batch_normalization_11 (Bat (None, 7, 7, 128) 512
ch归一化）
leaky_re_lu_15 (LeakyReLU) (无, 7, 7, 128) 0
dropout_7 (Dropout) (无, 7, 7, 128) 0
flatten_3（展平）（无，6272）0
dense_7（密集）（无，1）6273
================================================= ================ 总参数：139,073 可训练参数：138,817 不可训练的参数：256 ______________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________
我正在尝试更改 batch_size 但不起作用。我有 128*128 的图像用于训练模型。