The code is implemented with Python(3.6) and Pytorch(1.0.0).

To install the required python packages, run

pip install -r requirements.txt

Dataset and pretrained models could be calculated and found in UODTN, while we directly take them.

The source dataset could be found inhere

For the target dataset, one can download from here. The link is here

The image list of the dataset should be edit to the path of images in 'data/new_AwA2.txt' and 'data/WEB_3D3_2.txt'

The models could be calculated in UODTN, while we take the related models as follows:

-"base_net_pretrained_on_I2AwA2_source_only" is a trained feature extractor for I2AwA.

-"awa_50_cls_basic" contains 50 initial classifiers for AwA2.

These models should be put in the folder in "model/"

python train_loader.py --gpu_id 0

more options could be found in help

The codes are borrowed from UODTN