caposerenity / daps Goto Github PK

Hi, I have question about the implementation in hm.py loss.

Why did you use masked_softmax in the above code?

and also what is the meaning of below code?

nums = torch.zeros(labels.max()+1, 1).float().cuda()
nums.index_add_(0, labels, torch.ones(self.num_samples,1).float().cuda())
mask = (nums>0).float()
sim /= (mask*nums+(1-mask)).clone().expand_as(sim)

I think this code is same as SPCL re-ID code, but there is not much explantation.

Is using masked re-ID loss neccessary?

Hi, I am facing some error while running your codes.

There was some error preparing the setting
such as faiss-gpu, apex, ex..
but I made the code running and like below.

THe memory seems to be initalized but the error comes up right at the
begin of training.

Could you give some help?

In that part of the code, CUHK-SYSU specifies len_itr as 2240, but the train_log you uploaded shows 1420. Why is there such a difference?

Does "undetected person" mean someone who needs to find it but couldn't find it, or someone who doesn't have to? If it's "someone you don't have to find" does it mean someone who isn't ground truth labeling in the dataset?

Hi. Thank you for your great work.

I was wondering why is the LUT_SIZE=6015
when the target domain is set to CUHK-SYSU?

Also the CUHK-SYSU(11,206) has more scene images than PRW(5,704).
If the batch size is set to 4 (2 for CUHK and 2 for PRW),
in each epoch are there some images in CUHK-SYSU that are not fed into the model for training?