/DATE

Public code of AAAI 2023 paper "Discriminability and Transferability Estimation: A Bayesian Source Importance Estimation Approach for Multi-Source-Free Domain Adaptation"

Apache License 2.0Apache-2.0

DATE

Discriminability and Transferability Estimation

Overview

This repository is a PyTorch implementation of the paper.

Dataset

  • Manually download the datasets Office, Office-Home, Office-Caltech from the official websites.
  • Move gen_list.py inside data directory.
  • Generate '.txt' file for each dataset using gen_list.py (change dataset argument in the file accordingly).

Training

  • Train source models (shown here for Office with source A)
python train_source.py --dset office --s 0 --max_epoch 100 --trte val --gpu_id 0 --output ckps/source/
  • Adapt to target (shown here for Office with target D) based on SHOT
python adapt_multi_DATE_SHOT.py --dset office --t 1 --max_epoch 15 --gpu_id 0 --output_src ckps/source/ --output ckps/adapt