【Pytorch】实现subpixle上采样及下采样

最新推荐文章于 2024-04-15 22:10:12 发布

Shwan_Ma

最新推荐文章于 2024-04-15 22:10:12 发布

阅读量2.6k

点赞数 1

分类专栏： pytorch 文章标签： pytorch

本文链接：https://blog.csdn.net/shwan_ma/article/details/103604372

版权

Pytorch 实现subpixle上采样及下采样，类似与tensorflow的tf.depth_to_space , tf.space_to_depth

def shuffle_down(inputs, scale):
    N, C, iH, iW = inputs.size()
    oH = iH // scale
    oW = iW // scale

    output = inputs.view(N, C, oH, scale, oW, scale)
    output = output.permute(0,1,5,3,2,4).contiguous()
    return output.view(N, -1, oH, oW)

def shuffle_up(inputs, scale):
    N, C, iH, iW = inputs.size()
    oH = iH * scale
    oW = iW * scale
    oC = C // (scale ** 2)
    output = inputs.view(N, oC, scale, scale, iH, iW)
    output = output.permute(0,1,4,