python实现最大池化

实现一个二维矩阵的最大池化

import numpy as np

def max_pooling_2d(image, kernel_size, stride):
    # 获取输入图像的形状和卷积核大小
    input_height, input_width = image.shape
    kernel_height, kernel_width = kernel_size

    # 计算输出图像的大小
    output_height = (input_height - kernel_height) // stride + 1
    output_width = (input_width - kernel_width) // stride + 1

    # 创建一个与输出图像相同大小的零矩阵
    output = np.zeros((output_height, output_width))

    # 对输入图像进行最大池化
    for i in range(0, output_height):
        for j in range(0, output_width):
            # 计算池化窗口的位置
            row_start = i * stride
            row_end = row_start + kernel_height
            col_start = j * stride
            col_end = col_start + kernel_width

            # 提取池化窗口内的子图像
            sub_image = image[row_start:row_end, col_start:col_end]

            # 对子图像进行最大池化并将结果赋值给输出图像
            output[i, j] = np.max(sub_image)

    return output

# 示例用法
image = np.array([[1, 2, 3, 4],
                  [5, 6, 7, 8],
                  [9, 10, 11, 12],
                  [13, 14, 15, 16]])

kernel_size = (2, 2)
stride = 2

result = max_pooling_2d(image, kernel_size, stride)
print(result)