Implementation of "Adversarial Discriminative Domain Adaptation" in PyTorch
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
fnakamura db0012d698
Merge pull request #1 from Fujiki-Nakamura/wip/repro
6 years ago
input add .gitignore 6 years ago
outputs add .gitignore 6 years ago
.gitignore add .gitignore 6 years ago
README.md update README 6 years ago
adversarial.png update README 6 years ago
experiment.py logger 6 years ago
main.py update 6 years ago
models.py add batch norm 6 years ago
target_domain.png update README 6 years ago
train_source.py add logger 6 years ago
trainer.py [misc] refactor logs, save target model, tensorboard 6 years ago
utils.py logger 6 years ago

README.md

ADDA.PyTorch

implement Adversarial Discriminative Domain Adapation in PyTorch

Example

$ python train_source.py --logdir outputs
$ python main.py --logdir outputs --trained outputs/best_model.pt --slope 0.2

Result

SVHN -> MNIST

Paper This Repro
Source only 0.601 0.659
ADDA 0.760 ~0.83

adversarial target_domain

Resource