WPD: Weather prompt driven zero-shot adverse condition depth estimation

doi:10.1016/j.patcog.2025.112963

WPD: Weather prompt driven zero-shot adverse condition depth estimation

2025-12-26

0

PRE

AI

R

Rui Tian

张

张永强 (Yongqiang Zhang)

DOI：10.1016/j.patcog.2025.112963

Original

Original request for help

Save

Abstract

En 中文

• A novel approach WPD is proposed for zero-shot adverse condition depth estimation. • A novel Prompt Guided Affine Transformation optimizes the learnable affine transformation to estimate unseen target domain visual representations. • A novel Source-Target Visual Consistency is designed to preserve the content of the source-domain images. • Experiments show excellent performance on multiple datasets under night, rain, foggy and Twililight conditions.

WPD: Weather prompt driven zero-shot adverse condition depth estimation

Abstract

Journal

Organization