卷积神经网络中经卷积、池化后张量形状的变化

  • 'SAME':
    out_height = ⌈ float (in_height) / float (strides[1]) ⌉  (向上取整)
    out_width = ⌈ float (in_width) / float (strides[2]) ⌉ 
  • 'VALID':
    out_height = ⌈ float (in_height - filter_height + 1) / float (strides[1]) ⌉ 
    out_width = ⌈ float (in_width - filter_width + 1) / float (strides[2]) ⌉ 
发布了18 篇原创文章 · 获赞 15 · 访问量 3万+

猜你喜欢

转载自blog.csdn.net/Juicewyh/article/details/98739675