Under .src/model/mfsr.py
the function is blow
=============解释
def mc(self, tensor):
# mirror and then concat
inv_idx = torch.arange(tensor.size(0)-1, -1, -1).long().cuda()#img得到img倒序的索引 .long() 索引的记得转 不是应该在dim=w 镜像（水平翻转)吗 =======？？？？？？暂且认为是水平翻转？？？？ 问作者
inv_tensor = tensor.index_select(0, inv_idx)
inv_tensor = tensor[inv_idx]#得到相反顺序的tenor
return torch.cat([tensor, inv_tensor], 1)#在通道维度(dim=1)拼接：？？？岂不是两个不同的人组成不同的通道（-1,6,128,128） 不是应该在原来的水平面的吗
problem：why the input of encoder was preprocessed to be tensor with shape as（-1,6,128,128）which include two imgs from difference person，may you explain why 。 becuase my ability of coding is poor，I don‘t konw whether I had a incorrect。hope to reply。
Under .src/model/mfsr.py
the function is blow
=============解释
def mc(self, tensor):
# mirror and then concat
inv_idx = torch.arange(tensor.size(0)-1, -1, -1).long().cuda()#img得到img倒序的索引 .long() 索引的记得转 不是应该在dim=w 镜像（水平翻转)吗 =======？？？？？？暂且认为是水平翻转？？？？ 问作者
inv_tensor = tensor.index_select(0, inv_idx)
inv_tensor = tensor[inv_idx]#得到相反顺序的tenor
return torch.cat([tensor, inv_tensor], 1)#在通道维度(dim=1)拼接：？？？岂不是两个不同的人组成不同的通道（-1,6,128,128） 不是应该在原来的水平面的吗
problem：why the input of encoder was preprocessed to be tensor with shape as（-1,6,128,128）which include two imgs from difference person，may you explain why 。 becuase my ability of coding is poor，I don‘t konw whether I had a incorrect。hope to reply。