This repository implements the algorithm proposed in Unsupervised Domain Adaptation via Regularized Conditional Alignment, ICCV (2019) using PyTorch.
python 3.6
numpy 1.17.2
torch 1.4.0
torchvision 0.5.0
scikit-image
googledrivedownloader
requests
PyYAML
You can choose the source and the target dataset and run train.py:
python3 train.py --source_dataset=CIFAR --target_dataset=STL
python3 train.py --source_dataset=STL --target_dataset=CIFAR
python3 train.py --source_dataset=SVHN --target_dataset=MNIST
python3 train.py --source_dataset=MNIST --target_dataset=SVHN
python3 train.py --source_dataset=DIGIT --target_dataset=SVHN
python3 train.py --source_dataset=MNIST28 --target_dataset=MNISTM28