zifuwan
commented
5 months ago Hi, this is the deep supervision block we propose. We upsample all side outputs to the same size as the input image so as to calculate loss.
Open wusuowei666 opened 12 months ago
zifuwan
commented
5 months ago Hi, this is the deep supervision block we propose. We upsample all side outputs to the same size as the input image so as to calculate loss.
代码中的X是decoder中所有特征图的集合吗，最后的X的 分辨率是多少呢，为什么要用这个代码来上采样，而其余的都是规定好的分辨率
为什么要对X 上采样四倍呢，其余的分辨率不变
求告知！