► Keras 3 API 文档 / 层 API / 池化层 / MaxPooling3D层

MaxPooling3D层

`MaxPooling3D` class

keras.layers.MaxPooling3D(
    pool_size=(2, 2, 2),
    strides=None,
    padding="valid",
    data_format=None,
    name=None,
    **kwargs
)

最大池化操作,用于3D数据（空间或时空）.

沿其空间维度（深度、高度和宽度）对输入进行下采样,通过在输入窗口（大小由pool_size定义）中取每个输入通道的最大值.窗口沿每个维度按strides移动.

参数: pool_size: 整数或3个整数的元组,用于下采样的因子（dim1, dim2, dim3）.如果只指定一个整数,则所有维度将使用相同的窗口长度. strides: 整数或3个整数的元组,或None.步幅值.如果为None,则默认为pool_size.如果只指定一个整数,则所有维度将使用相同的步幅大小. padding: 字符串,"valid"或"same"（不区分大小写）."valid"表示无填充."same"结果是在输入的左右或上下均匀填充,使得输出具有与输入相同的高度/宽度维度. data_format: 字符串,"channels_last"或"channels_first".输入中维度的顺序."channels_last"对应于形状为(batch, spatial_dim1, spatial_dim2, spatial_dim3, channels)的输入,而"channels_first"对应于形状为(batch, channels, spatial_dim1, spatial_dim2, spatial_dim3)的输入.它默认为在~/.keras/keras.json中的image_data_format值.如果你从未设置它,则它将是"channels_last".

输入形状:

如果data_format="channels_last": 形状为(batch_size, spatial_dim1, spatial_dim2, spatial_dim3, channels)的5D张量
如果data_format="channels_first": 形状为(batch_size, channels, spatial_dim1, spatial_dim2, spatial_dim3)的5D张量

输出形状:

如果data_format="channels_last": 形状为(batch_size, pooled_dim1, pooled_dim2, pooled_dim3, channels)的5D张量
如果data_format="channels_first": 形状为(batch_size, channels, pooled_dim1, pooled_dim2, pooled_dim3)的5D张量

示例:

depth = 30
height = 30
width = 30
channels = 3

inputs = keras.layers.Input(shape=(depth, height, width, channels))
layer = keras.layers.MaxPooling3D(pool_size=3)
outputs = layer(inputs)  # 形状: (batch_size, 10, 10, 10, 3)

MaxPooling3D层

MaxPooling3D class

MaxPooling3D层

MaxPooling3D class

`MaxPooling3D` class