Awesome

BTDA

implement for paper Rethinking Domain Adaptation Blending target Domain Adaptation by Adversarial Meta Adaptation Network

Requirements

• python 2.7
• PyTorch 0.4.1
• cuda 8.0
• numpy
• tqdm
• visdom
• PIL
• argparse
• sklearn
• torchvision

download dataset

cd BTDA

bash download_dataset.sh

digit dataset

cd BTDA/Digit

python train.py --source_name mnist --gpu_id 0

office31 dataset

cd BTDA/Office

python alexnet_train.py --dataset_name Office31 --source_name amazon --gpu_id 0

python resnet_train.py --dataset_name Office31 --source_name amazon --gpu_id 0

officeHome dataset

cd BTDA/Office

python alexnet_train.py --dataset_name OfficeHome --source_name A --gpu_id 0

python resnet_train.py --dataset_name OfficeHome --source_name A --gpu_id 0

some result modles

alexnet amazon->webcam,dslr acc:74.6 https://drive.google.com/open?id=1_6KGKGfhNJkOo5j0p154FQKCIavrrrpr

resnet amazon->webcam,dslr acc:90.1 https://drive.google.com/open?id=1UGMjGaaHKQvEAGLGym0R-fQyQIeOkjOa