如果我这三张图片已经加载到网络中，维度是[1,224,224,3]，那代码要怎么改呢

如果你已经将这三张图片加载到网络中，维度是`[1,224,224,3]`，那么可以按照以下方式修改代码：

def compute_perceptual_loss(style_features, content_features, generated_features):
    # 计算风格损失
    style_loss = 0.
    for style_feature, generated_feature in zip(style_features, generated_features):
        style_gram = gram_matrix(style_feature)
        generated_gram = gram_matrix(generated_feature)
        style_loss += tf.reduce_mean(tf.square(style_gram - generated_gram)) / (4. * tf.cast(tf.size(generated_feature), tf.float32) ** 2)

    # 计算内容损失
    content_loss = tf.reduce_mean(tf.square(content_features - generated_features))

    # 计算感知损失（即风格损失和内容损失之和）
    perceptual_loss = style_loss + content_loss

    return perceptual_loss.numpy()

# 加载风格图片、内容图片和生成图片
style_image = ...  # 维度为[1,224,224,3]
content_image = ...  # 维度为[1,224,224,3]
generated_image = ...  # 维度为[1,224,224,3]

# 加载VGG19模型，并指定需要输出的层
vgg19 = tf.keras.applications.VGG19(include_top=False, weights='imagenet')
style_layers = ['block1_conv1', 'block2_conv1', 'block3_conv1', 'block4_conv1']
content_layers = ['block5_conv2']

# 构建模型，用于获取风格图片、内容图片和生成图片在指定层的特征
outputs = [vgg19.get_layer(name).output for name in style_layers + content_layers]
model = tf.keras.models.Model(inputs=vgg19.inputs, outputs=outputs)

# 计算风格图片、内容图片和生成图片在指定层的特征
style_features = model(style_image)[:len(style_layers)]
content_features = model(content_image)[len(style_layers):]
generated_features = model(generated_image)[len(style_layers):]

# 计算感知损失
perceptual_loss = compute_perceptual_loss(style_features, content_features, generated_features)

在代码中，`style_image`、`content_image`和`generated_image`分别表示风格图片、内容图片和生成图片，维度均为`[1,224,224,3]`。首先，加载VGG19模型并指定需要输出的层，然后构建模型，用于获取风格图片、内容图片和生成图片在指定层的特征。接着，计算风格图片、内容图片和生成图片在指定层的特征，最后计算感知损失（即风格损失和内容损失之和）。