Home

Awesome

PyTorch-ADDA

A PyTorch implementation for Adversarial Discriminative Domain Adaptation.

Environment

Usage

I only test on MNIST -> USPS, you can just run the following command:

python3 main.py

Network

In this experiment, I use three types of network. They are very simple.

Result

MNIST (Source)USPS (Target)
Source Encoder + Source Classifier99.140000%83.978495%
Target Encoder + Source Classifier97.634409%

Domain Adaptation does work (97% vs 83%).