futabasakuraxd/farewell-to-mutual-information-variational-distiilation-for-cross-modal-person-re-identification

farewell-to-mutual-information-variational-distiilation-for-cross-modal-person-re-identification's Issues

关于VSD中KL散度计算的一点疑问

作者您好，您在VSD中计算了p(y|v)和p(y|z)的散度，计算的代码如下：

 vsd_loss = kl_div(input=self.softmax(i_observation[0].detach() / self.args.temperature),
                          target=self.softmax(i_representation[0] / self.args.temperature))

但是根据我的理解，p(y|v)和p(y|z)的KL散度应该是如下计算：

vsd_loss = kl_div(input=torch.nn.LogSoftmax(dim=1)(i_representation[0] / self.args.temperature),
                          target=self.softmax(i_observation[0].detach() / self.args.temperature))

请问为什么您在代码中是使用第一种实现方式？这两种实现方式是等价的吗？

model issue

Could you share your "model_best.pth"? Where can we find them?

the implementation is quite different from the paper

In the code, I notice that you use the grayscale images as an input to Ei. Isn't it supposed to be infrared images?
and why did you mix the infrared and RGB images then fed it as an input to Ev?
and you didn't even use VML loss. There is significant difference between the code and the paper.

Could you please clarify the reason behind the difference.

Thank you in advance!

网络结构

您好，请问，网络结构那里，可见光分支的输入是一组可见光和一组红外图像，然后红外分支那里是这一组图像的灰度图像是吗？

REGDB数据的整理

RegDB.py中# 412 identities with 3 camera views each，您用的数据集是跟SYSU-MM01那样分了几个cam的吗，最后是将RegDB数据集的所有图片汇总到一个images文件夹里面吗，images文件夹中是没有其他的子文件夹吗

update