Pytoch_DANN

This is a implementation of Domain-Adversarial Training of Neural Networks
with pytorch. This paper introduced a simple and effective method for accompli-
shing domian adaptation with SGD with a GRL(Gradient Reveral Layer). According
to this paper, domain classifier is used to decrease the H-divergence between
source domain distribution and target domain distribution. For the tensorflow
version, you can see tf-dann.

requirements

python3.6.2
pip install -r requirements.txt

Data

In this work, MNIST and MNIST_M datasets are used in experiments. MNIST dataset
can be downloaded with torchvision.datasets. MINIST_M dataset can be downloa-
ded at Yaroslav Ganin's homepage or direct Google Drive link. Then you can extract the file to your data dire-
ctory and run the preprocess.py to make the directory able to be used with
torchvision.datasets.ImageFolder:

# Download MNIST_M dataset to data folder 
cd data

## unzip the files
tar -zvxf ./mnist_m.tar.gz
# IF the above command doesn't work use the following
tar -vxf mnist_m.tar.gz

# To process the MNIST_M dataset
python3 ./util/preprocess.py

# To Download the MNIST dataset just run the main.py script
python3 main.py

Experiments

You can run main.py to implements the MNSIT experiments for the paper with the
similar model and same paramenters.The paper's results and this work's results a-
re as follows:

Method	Target Acc(paper)	Target Acc(this work)
Source Only	0.5225	0.5189
DANN	0.7666	0.7600

Experiment on SVHN->MNIST is added in this project, but some bugs are not fixed.
The accuracies of source and target domains are not good at the same time.

Experiment on SynDig->SVHN is added.