hongbin98 / proca Goto Github PK

Hi!
First of all, thank you for your nice work!
While I read your paper and code in github, I have a question about initialization for target domain model.

I saw that the training process of other works in Unsupervised Domain Adaptation follows 1) Pre-train a model with source domain 2) Initialize target model with source domain pre-trained weight.

But in your code, it is hard to find Initialization step with source domain pre-trained weight.

Is there any initialization stage in the code, or am I mistaken?

I'm interested in this work and I want to know if it is possible in UDA seg

Could you please share the code of CIDA method?
Thank you!

As the readme said that the ImageNet dataset split is too large to update, I am already emailed you guys, or I can get this file from other sources?

I use the OH_adapt_2_target.py to train the pretrain model on the office-31 dataset, but I cannot reach the result accuracy which showing in the paper, is there any modification should apply on this file to make it suit for office-31 dataset.