Generalized Domain Conditioned Adaptation Network

The GDCAN and DCAN algorithms implemented in Pytorch.

(AAAI2020 - DCAN) Domain Conditioned Adaptation Network
(T-PAMI - GDCAN) Generalized Domain Conditioned Adaptation Network

Introduction

We relax a shared-convnets assumption made by previous DA methods and propose a Domain Conditioned Adaptation Network (DCAN), which aims to excite distinct convolutional channels with a domain conditioned channel attention mechanism. As a result, the critical low-level domain-dependent knowledge could be explored appropriately. Moreover, to effectively align high-level feature distributions across two domains, we further deploy domain conditioned feature correction blocks after task-specific layers, which will explicitly correct the domain discrepancy.

Citation

If you find this work valuable or use our code in your own research, please consider citing us with the following bibtex:

@inproceedings{li20DCAN,
    title = {Domain Conditioned Adaptation Network},
    author = {Li, Shuang and Liu, Chi Harold and Lin, Qiuxia and Xie, Binhui and Ding, Zhengming and Huang, Gao and Tang, Jian},
    booktitle = {Thirty-Fourth AAAI Conference on Artificial Intelligence},    
    pages     = {11386--11393},
    publisher = {{AAAI} Press},
    year = {2020}
}
@article{li2021generalized,
  author          = {Li, Shuang and Xie, Binhui and Lin, Qiuxia and Liu, Chi Harold and Huang, Gao and Wang, Guoren},
  title           = {Generalized Domain Conditioned Adaptation Network},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  volume          = {},
  number          = {},
  pages={1-1},
  year            = {2021},
  doi={10.1109/TPAMI.2021.3062644}
}

BIT-DA/GDCAN

Generalized Domain Conditioned Adaptation Network

Introduction

Citation