Improving Semi Supervised Domain Adaptation using Target Selection and Semantics

Install

conda env create -n SSDA.yml

The code is written for Pytorch 0.4.0, but should work for other version with some modifications.

Data preparation (DomainNet)

Download the cleaned version of the domainnet data from here and place them inside the './data/multi/' folder.

The images will be stored in the following way.

./data/multi/real/category_name,

./data/multi/sketch/category_name

The dataset split files are stored as follows,

./data/txt/multi/labeled_source_images_real.txt,

./data/txt/multi/unlabeled_target_images_sketch_3.txt,

With regard to office and office home dataset, store the image files in the following ways,

./data/office/amazon/category_name, ./data/office_home/Real/category_name,

We provide the split of office and office-home.

Training

To run training using alexnet,

sh run_train.sh gpu_id method alexnet

where, gpu_id = 0,1,2,3...., method=[MME,ENT,S+T].

Reference

This implementation is based on the base MME implementation from Kuniaki Saito and Donghyun Kim