Domain Adaptation via Surrogate Mutual Information Maximization

Introduction

This is the code of our work Domain Adaptation via Surrogate Mutual Information Maximization published on IJCAI 2022.

To reproduce our experiments, please first install conda environment and download datasets.

Environment

conda create -n sida python=3.6
source activate sida

pip install -r requirements.txt

Datasets

cd experiments
cd dataset

# download Office-31 from https://drive.google.com/file/d/0B4IapRTv9pJ1WGZVd1VDMmhwdlE/view?resourcekey=0-gNMHVtZfRAyO_t2_WrOunA
# download Office-Home from https://drive.google.com/file/d/0B81rNlvomiwed0V1YUxQdC1uOTg/view?usp=sharing&resourcekey=0-2SNWq0CDAuWOBRRBL7ZZsw
# download VisDA2017 from
# https://drive.google.com/file/d/0BwcIeDbwQ0XmdENwQ3R4TUVTMHc/view?usp=sharing
# https://drive.google.com/file/d/0BwcIeDbwQ0XmUEVJRjl4Tkd4bTA/view?usp=sharing
# https://drive.google.com/file/d/0BwcIeDbwQ0XmdGttZ0k2dmJYQ2c/view?usp=sharing

bash data_preprocess.sh

cd ..
cd ..

Run Experiments

Run train.py with arguments --cfg to the configuration file and --exp_name for the output directory. The experiment log file and the checkpoints will be stored at ./experiments/ckpt/${experiment_name}

For example, to run experiments of OfficeHome:

CUDA_VISIBLE_DEVICES=0  python train.py --cfg ./experiments/config/OfficeHome/office31_train_A2C_cfg.yaml  --exp_name officehome_a2c

for VisDA-2017:

CUDA_VISIBLE_DEVICES=0  python train.py --cfg ./experiments/config/OfficeHome/visda17_train_train2val_cfg.yaml  --exp_name visda2017 

and for Office-31:

CUDA_VISIBLE_DEVICES=0  python train.py --cfg experiments/config/Office-31/office31_train_amazon2dslr_cfg.yaml  --exp_name office31_a2d 

Acknowledgement

The clustering framework of our code references https://github.com/kgl-prml/Contrastive-Adaptation-Network-for-Unsupervised-Domain-Adaptation.