asahi417 / deepdomainadaptation

Tensorflow deep learning based domain adaptation model implementations with experiment of estimate MNIST by SVHN data (SVHN -> MNIST): DANN (domain-adversarial neural network), Deep JDOT (joint distribution optimal transportation)

Deep Domain Adaptation

Tensorflow implementation of deep learning based domain adaptation models. See more in implemented algorithm.

Work in progress: Experiment results will be appeared soon!

Get started

git clone https://github.com/asahi417/DeepDomainAdaptation
cd DeepDomainAdaptation
pip install .

• install error
  You might have install error if your environment dosen't have numpy and cython, due to the pot library. Then, import them before install this repo.

pip install numpy
pip install cython
pip install .

Script

bin/script_tfrecord.py

This script converts dataset to tfrecord format.

usage: script_tfrecord.py [-h] --data [DATA]

optional arguments:
  -h, --help     show this help message and exit
  --data [DATA]  dataset name in dict_keys(['mnist', 'svhn'])

bin/script_train.py

This script is to train models.

usage: script_train.py [-h] -m [MODEL] -e [EPOCH] [-v [VERSION]]

optional arguments:
  -h, --help            show this help message and exit
  -m [MODEL], --model [MODEL]
                        Model name in dict_keys(['dann', 'deep_jdot'])
  -e [EPOCH], --epoch [EPOCH]
                        Epoch
  -v [VERSION], --version [VERSION]
                        Checkpoint version if train from existing checkpoint

List of Models

• Domain Adversarial Neural Network
  • Ganin, Yaroslav, et al. "Domain-adversarial training of neural networks." The Journal of Machine Learning Research 17.1 (2016): 2096-2030.
  • implementation
• Deep Joint Distribution Optimal Transport
  • Damodaran, Bharath Bhushan, et al. "DeepJDOT: Deep Joint distribution optimal transport for unsupervised domain adaptation." arXiv preprint arXiv:1803.10081 (2018).
  • Nicolas Courty, et al. "Joint distribution optimal transportation for domain adaptation" Advances in Neural Information Processing Systems. 2017.
  • implementation