[Task]: patchcoreensemble

[2324773711]

What is the motivation for this task?

How are the features of different backbone networks combined in PatchCoreEmSemble? cancat or add or what?

Describe the solution you'd like

no solution!

Additional context

No response

[samet-akcay]

Can you elaborate a bit more? Which PatchCoreEmSemble model are you interested in specifically? Maybe some links could help?

[2324773711]

https://arxiv.org/pdf/2106.08265v2--PatchCore paper Table 4, PatchCoreEmsemble there are three Backbone.DenseN-201&RNext-101&WRN-101.How did they combination.like concat or add,or other ways?

[abc-125]

I think they just append it to the list of all features, but you can wait for the reply in the original repository to be sure.

[alexriedel1]

They minmax scale the scores and take the mean after
https://github.com/amazon-science/patchcore-inspection/blob/fcaa92f124fb1ad74a7acf56726decd4b27cbcad/bin/run_patchcore.py#L114-L118

[2324773711]

步步 ***@***.***  My question is how the features of the three different backbone(DenseN-201, RNext-101, WRN-101) networks in Table 4 are combined in the ablation experiment of the PatchCore paper.

…

------------------ 原始邮件 ------------------ 发件人: "openvinotoolkit/anomalib" ***@***.***>; 发送时间: 2024年6月25日(星期二) 下午5:41 ***@***.***>; ***@***.******@***.***>; 主题: Re: [openvinotoolkit/anomalib] [Task]: patchcoreensemble (Discussion #2149) They minmax scale the scores and take the mean after https://github.com/amazon-science/patchcore-inspection/blob/fcaa92f124fb1ad74a7acf56726decd4b27cbcad/bin/run_patchcore.py#L114-L118 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>