Home

Awesome

Dual-Path Adversarial Lifting for Domain Shift Correction in Online Test-time Adaptation

This is the official project repository for Dual-Path Adversarial Lifting for Domain Shift Correction in Online Test-time Adaptation. This repository is built based on the SAR 🔗.

<p align="center"> <img src="figures/DPAL.png" alt="DPAL" width="100%" align=center /> </p>

Installation:

DPAL depends on

This repository contains code for evaluation on ImageNet-C, ImageNet-R, ImageNet-A, and VisDA-2021 with ViT. The pre-trained ViT-B model weights are from timm.

Usage:

bash run.sh

Acknowledgment

The code is inspired by the Tent 🔗 and SAR 🔗.