STFAR: Test-time adaptive object detection through self-training and feature alignment regularization

Nanqing Liu

doi:10.1016/j.eswa.2026.132594

STFAR: Test-time adaptive object detection through self-training and feature alignment regularization

2026-04-28

0

PRE

AI

DOI：10.1016/j.eswa.2026.132594

Original

Original request for help

Save

Abstract

En 中文

• TTAOD enhances detector robustness against unknown corruptions at test time. • A self-training framework with feature alignment enables stable test-time adaptation. • GFA aligns features across domains; PFC improves foreground feature separability. • Benchmarks for TTAOD are built with code released for reproducibility.

Keywords:

TTAOD

self-training

feature alignment

object detection

test-time adaptation

STFAR: Test-time adaptive object detection through self-training and feature alignment regularization

Abstract

Journal

Organization