Code release for "Meta-reweighted Regularization for Unsupervised Domain Adaptation"
This work proposes a novel regularization mechanism in deep UDA via meta-learning to guide the classifier to adapt better to the target domain.
VisDA-2017 dataset can be found here in the classification track.
Run the following command in shell:
visda-2017 DANN+MetaReg
CUDA_VISIBLE_DEVICES=0 python MetaReg_dann.py /data1/TL/data/ -d VisDA2017 -s T -t V -a resnet50 -j 8 --epochs 30 --i 1000 --log_filename metareg-dann-visda --seed 2020
visda-2017 CDAN+MetaReg
CUDA_VISIBLE_DEVICES=0 python MetaReg_cdan.py /data1/TL/data/ -d VisDA2017 -s T -t V -a resnet50 -j 8 --epochs 30 --i 1000 --log_filename metareg-dann-visda --seed 2020
Some codes in this project are borrowed from Transfer-Learn and L2RW. We thank them for their excellent projects.
If you have any problem about our code, feel free to contact
or describe your problem in Issues.