CDAN

Code release for "Conditional Domain Adversarial Network" (NIPS 2018)

Prerequisites

• PyTorch >= 0.4.0 (with suitable CUDA and CuDNN version)
• torchvision >= 0.2.1
• Python3
• Numpy
• argparse
• PIL

Dataset

Digits

Processed SVHN_dataset is here. We change the original mat into images. Other transformed images are in data/svhn2mnist and data/usps2mnist. Dataset_train.txt are lists for source and target domains and Dataset_test.txt are lists for test.

Office-31

Office-31 dataset can be found here.

Office-Home

Office-Home dataset can be found here.

VisDA-2017

VisDA 2017 dataset can be found here in the classification track.

Image-clef

We release the Image-clef dataset we used here.

Training

All the parameters are set to optimal in our experiments. The following are the command for each task. The test_interval can be changed, which is the number of iterations between near test.

SVHN->MNIST
python train_svhnmnist.py --gpu_id id --epochs 50

USPS->MNIST
python train_uspsmnist.py --gpu_id id --epochs 50 --task USPS2MNIST

MNIST->USPS
python train_uspsmnist.py --gpu_id id --epochs 50 --task MNIST2USPS

Office-31

pythonn train_image.py --gpu_id id --net ResNet50 --dset office --test_interval 500 --s_dset_path ../../data/office/amazon_list.txt --t_dset_path ../../data/office/webcam_list.txt

Office-Home

pythonn train_image.py --gpu_id id --net ResNet50 --dset office-home --test_interval 2000 --s_dset_path ../../data/office-home/Art.txt --t_dset_path ../../data/office-home/Clipart.txt

VisDA 2017

pythonn train_image.py --gpu_id id --net ResNet50 --dset visda --test_interval 5000 --s_dset_path ../../data/visda-2017/train_list.txt --t_dset_path ../../data/visda-2017/validation_list.txt

Image-clef

pythonn train_image.py --gpu_id id --net ResNet50 --dset image-clef --test_interval 500 --s_dset_path ../../data/image-clef/b_list.txt --t_dset_path ../../data/image-clef/i_list.txt

If you want to run the random version of CDAN, add --random as a parameter.

Citation

If you use this code for your research, please consider citing:

@inproceedings{long2018conditional,
  title={Conditional adversarial domain adaptation},
  author={Long, Mingsheng and Cao, Zhangjie and Wang, Jianmin and Jordan, Michael I},
  booktitle={Advances in Neural Information Processing Systems},
  pages={1645--1655},
  year={2018}
}

Contact

If you have any problem about our code, feel free to contact

• caozj@cs.stanford.edu
• longmingsheng@gmail.com

or describe your problem in Issues.