Using the same target dataset for both training and testing in Office31

[bardisafa]

Hi, I realized that you are testing the model on the same target dataset that was used during trainng and active sampling ('tr_or_te' doesn't exist in the filename). Here:

ActiveLearning-SDM/run.py

Lines 29 to 32 in 0ee700e

	train_target_data = loader.get_data(cfg.DATA_NAME,os.path.join(cfg.DATA_PATH,cfg.DATA_TARGET),
	transform.test_transform,tr_or_te='train',n_views=cfg.N_VIEWS)
	test_target_data = loader.get_data(cfg.DATA_NAME,os.path.join(cfg.DATA_PATH,cfg.DATA_TARGET),
	transform.test_transform,tr_or_te='test',n_views=cfg.N_VIEWS)

Could you explain why you did that? Thanks

[lyxok1]

Hi, we train and test the model in the same target dataset to follow a standard setting the field of Domain Adaptation, you can refer to some previous work like TQS or CDAN, these classic work also take target domain as available data without label during training.