ZhangJUJU / Open-set-domain-adaptation

Open set domain adaptation code DAOD

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

  1. Implement_XXXX is the main text. All parameters have been tuned and fixed. One can implement code directly after changing the path.

  2. The features of Office-31, Office-Home and Imageclef-DA are extracted from deep models (pretrained without finetune).

  3. For all VGG features, we use the following data preprocessing method:

    Xs = Xs ./ repmat(sum(Xs,2),1,size(Xs,2));

    Xt = Xt ./ repmat(sum(Xt,2),1,size(Xt,2));

    Xs=normr(zscore(Xs));

    Xt=normr(zscore(Xt));

    The preprocessing method follows JGSA https://documents.uow.edu.au/~jz960/.

  4. The original code is written using Matlab R2019a.

  5. There are 98 open set domain adaptation tasks.

About

Open set domain adaptation code DAOD


Languages

Language:MATLAB 100.0%