- Python 3.7
- torchvision 0.9.0
- PyTorch 1.8.0
- Unsupervised DA on
Office31
,OfficeHome
, andVisDA2017
datasets:sh runUDA.sh
- Unsupervised DA on
ImageNet-scale
dataset:python main.py --dataset_root ./data/ --src IN --tgt INR --contrast_dim 256 --module domain_loss --cw 1 --lr 0.003 --batch_size 32 --max_key_size 20 --max_iterations 50000
- Semi-supervised DA on
COVID-19
dataset:sh runSSDA.sh
-
The training log will be generated in the folder with
--log_dir
. We can visualize the training process throughtensorboard
as follows.tensorboard --logdir=/log_dir/ --host= `host address`
-
We uploaded the file
PythonGraphPers_withCompInfo.so
for computing theconnected components
. If you need to generate it, you can compile the C++ code in folderref
, run./compile_pers_lib.sh
(by default it requires Python 3.7. If you are using other Python versions, modify the command insidecompile_pers_lib.sh
). -
pybind11
is a lightweight header-only library that exposes C++ types in Python and vice versa, mainly to create Python bindings of existing C++ code. Our code will be further improved to make it cleaner and easier to use.
Note: Place the datasets in the corresponding data path.