tutorial.md

• This is the tutorial for reproducing the results in the paper, including the training and testing process.
• We recommend you to use the docker image to run the codes, which can avoid the environment problems.

Docker

Please ensure that you have installed the docker via the following command:

docker --version 

hai datasets info  # View all datasets available
hai datasets download SFID  # download SFID. You can specify version by DATASET_NAME:VERSION, e.g. SFID:latest
unzip SFID.zip
rm SFID.zip

The dataset is unzipped to the SFID folder, and the training logs in the runs folder.