search

yogeshbalaji / Generate_To_Adapt

Implementation of "Generate To Adapt: Aligning Domains using Generative Adversarial Networks"
https://arxiv.org/pdf/1704.01705.pdf
142 stars 33 forks source link

The implementation of source only is a bit confusing #3

Closed zhaoxin94 closed 6 years ago

zhaoxin94 commented 6 years ago

The implementation of source_only is a bit confusing.

In my opinion, source_only should be trained on the source domain and tested on the target domain when no adaptation is performed.

Source only corresponds to the lower performance bound

yogeshbalaji commented 6 years ago

Yes. What you said is right. If you look at the Sourceonly class in trainer.py, the model is trained only on the source domain data. There is no adaptation performed here. I don't understand what the confusion is. Can you please elaborate more about your concern.

zhaoxin94 commented 6 years ago

Sorry, your code is completely correct！ I was wrong. I mistaken the validation set and the test set.

Thank you very much for your good job. It's helpful to me.