search

YanzuoLu / MLNet

[AAAI 2024] MLNet: Mutual Learning Network with Neighborhood Invariance for Universal Domain Adaptation
MIT License
12 stars 1 forks source link

实验设置疑惑 #1

Open happy-charity opened 3 months ago

happy-charity commented 3 months ago

作者你好，请问论文题目所说的UniDa是指在已知两域类别情况下分别做OPDA,CDA,ODA,PDA的实验来证明通用性吗？而不是真正在目标域类别信息未知的前提下判断是否为公共类还是私有类。 image

YanzuoLu commented 3 months ago

The training in UniDA has no information of the relation of label set between domains. Since this is domain adaptation so we will have the class labels of the source domain. The only things we have about the target domain are unlabelel images.