What's the neighbourhood size In PathCore implementations?

[mjack3]

Hello and thank for your contribution. Really good job

Trying your code i am wondering if this attribute n_reweight is the neighbourhood size described in the Paper.

[rvorias]

Hi,

No, the neighbourhood size is defined at

ind_knn_ad/indad/models.py

Line 241 in 57de628

self.average = torch.nn.AvgPool2d(3, stride=1)

We use the averaging to combined information of neighbouring patches. In this case the neighbourhood kernel is of size 3.

[mjack3]

Hello, thanks for your answer. I did a mistake in my question, I would like to where is defined the value of "b nearest patch-features"

[rvorias]

Yes, n_reweight = b in the paper.

[mjack3]

Thanks! to close this issue, I would appreciate to do a last question. In README section I see this about PatchCore

* actually does not have any knn mechanism, but shares many things implementation-wise.

is KNN implemented? If not, maybe I could make a develop

[rvorias]

That is actually a note on Padim, not PatchCore.