oh-yu / pytorch_dan

An implementation of Deep Adaptation Network with pytorch

Pytoch_DAN

This is a simple implementation of Learning Transferable Features with Deep Adaptation
Networks with pytorch. This paper introduced a simple and effective method for
accomplishing domian adaptation with MMD loss. According to this paper,
multi-layer features are adapted with MMD loss. In this paper, model is based
on AlexNet and tested on several datasets, while this work just utilizes
LeNet and tests on MNIST and MNIST_M datasets. The original implementation
in caffe is here.

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. 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:

python preprocess.py

If you could not download MNIST_M dataset from Yaroslav Ganin's homepage, you cou-
ld download it from MEGA Cloud. Once you download it, then you just need to unzip
the file to /data and the preprocess.py should not be used.

Experiments

You can run main.py to implements the MNSIT experiments. This work's results
are as follows:

Method	Target Acc(this work)
Source Only	0.5189
DAN	0.5829