qinghaoli0421/ps_code

PS

This repository contains the implementation in Pytorch for method PS of SFDA method.

Environment Requirement

• python 3.6.5
• pytorch 1.6.0
• torchvision 0.7.0
• cuda 10.1
• numpy, sklearn, scipy

Data

We use VisDA dataset in this code. Please download the dataset at VisDA dataset from the official websites, and modify the path of images in each '.txt' under the folder './data/'.

Training

Firstly, you need to train the source model by running this code

python train_source.py

Then, you can run the code of PS to perform SFDA,

python adaptation.py