我在配置 Keras/tensorflow 模型时遇到了困难

Question

我正在尝试创建这个可以帮助我生成、修改视频的人工智能模型。我以前遇到过一些错误，但我能够绕过它们。不管怎样，现在我陷入了一个可怕的、烦人的错误中，我在试图解决它的时候头疼，我累了，我真的需要一些帮助please:

错误说 X 样本（输入数组）=16 和 Y 样本（目标数组）=32 的大小不相等，这是完整的错误：

这是我的完整代码：

import os
import cv2
import numpy as np
import tensorflow as tf
from tensorflow.keras.layers import Dense, Reshape, BatchNormalization, Conv2DTranspose, Conv2D, LeakyReLU, Flatten
from tensorflow.keras.models import Sequential
from tensorflow.keras.optimizers import Adam

processed_data_folder = 'data/processed_data/'
model_folder = 'models/'
image_size = 64
batch_size = 16
epochs = 50
learning_rate = 0.0002
beta_1 = 0.5

def build_generator():
    model = Sequential()

    model.add(Dense(4 * 4 * 512, input_shape=(100,)))
    model.add(Reshape((4, 4, 512)))
    model.add(BatchNormalization())
    model.add(Conv2DTranspose(256, kernel_size=5, strides=2, padding='same', activation='relu'))
    model.add(BatchNormalization())
    model.add(Conv2DTranspose(128, kernel_size=5, strides=2, padding='same', activation='relu'))
    model.add(BatchNormalization())
    model.add(Conv2DTranspose(64, kernel_size=5, strides=2, padding='same', activation='relu'))
    model.add(BatchNormalization())
    model.add(Conv2DTranspose(3, kernel_size=5, strides=2, padding='same', activation='tanh'))

    return model

def build_discriminator():
    model = Sequential()

    model.add(Conv2D(64, kernel_size=5, strides=2, padding='same', input_shape=(image_size, image_size, 3)))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Conv2D(128, kernel_size=5, strides=2, padding='same'))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Conv2D(256, kernel_size=5, strides=2, padding='same'))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Conv2D(512, kernel_size=5, strides=2, padding='same'))
    model.add(LeakyReLU(alpha=0.2))
    model.add(Flatten())
    model.add(Dense(1, activation='sigmoid'))

    return model

generator = build_generator()
discriminator = build_discriminator()

discriminator.compile(loss='binary_crossentropy', optimizer=Adam(lr=learning_rate, beta_1=beta_1), metrics=['accuracy'])

discriminator.trainable = False
gan = Sequential([generator, discriminator])
gan.compile(loss='binary_crossentropy', optimizer=Adam(lr=learning_rate, beta_1=beta_1))

X_train = []
for video_folder in os.listdir(processed_data_folder):
    frames_folder = os.path.join(processed_data_folder, video_folder, 'frames')
    frames = []
    for frame_file in os.listdir(frames_folder):
        frame_path = os.path.join(frames_folder, frame_file)
        frame = cv2.imread(frame_path)
        frame = cv2.resize(frame, (image_size, image_size))
        frames.append(frame)
    X_train.append(frames)
X_train = np.array(X_train)

X_train = np.reshape(X_train, (-1, image_size, image_size, 3))
print(X_train.shape)

X_train = X_train / 127.5 - 1.

def generate_noise(batch_size, noise_shape):
    return np.random.normal(0, 1, size=(batch_size, noise_shape))

for epoch in range(epochs):

    noise = generate_noise(batch_size, 100)

    generated_images = generator.predict(noise)

    idx = np.random.randint(0, X_train.shape[0], batch_size)
    
    real_images = X_train[idx]

    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(generated_images, fake_labels)

    X = np.concatenate([real_images, generated_images])
    y = np.concatenate([real_labels, fake_labels])

    g_loss = gan.train_on_batch(noise, y)

    print(f"Epoch {epoch+1}/{epochs} [D real loss: {d_loss_real[0]:.4f}, D fake loss: {d_loss_fake[0]:.4f}, G loss: {g_loss:.4f}]")
    
    discriminator.trainable = True 
    
    generator_history = generator.fit(X_train, np.ones((X_train.shape[0], 1)), batch_size=batch_size, epochs=10, shuffle=True) 
    
    discriminator_history = discriminator.fit(X_train, y_train, batch_size=batch_size, epochs=10, shuffle=True)
    
generator.save(os.path.join(model_folder, 'video_generator.h5'))

批量大小为 32，由于系统内存不足，我不得不减少。

基本上我尝试了所有方法，从将帧图像大小调整为 32 再到 16，将批量大小从 64 减少到 32 再到 16。将帧限制为 1000 帧....很多方法和更新...我无法记住。