logo
CanyamAI summaries for academic research
language
MobileMobile
  • Phone FeatureWeChat Mini Program
EnglishEnglish
noticeNotice
BlogBlog
delete
arrow
Return

CrysFormer++: Dual-phase refinement learning for transparent object depth estimation

delete2025-10-17
delete0
PRE
AI
X
Xiaomei Zhang
邓敏 cover
邓敏 (Min Deng)
J
Jiwei Hu *
X
Xiao Huang
靳
靳淇文 (Qiwen Jin)
DOI：10.1016/j.eswa.2025.130043delete
deleteOriginal
deleteOriginal request for help
deleteShare
deleteSave
Abstract

Abstract

En 中文
• A hierarchical visual architecture with multi-scale gated fusion and fine-grained matching is proposed, coupling Mamba-Transformer features via dynamic gating and aligning spatial-semantic features to boost geometric fusion and structural representation. • A jointly optimized multi-constraint loss function constrains depth precision, spatial smoothness, confidence consistency, incorporates gated feature interaction and cross-scale constraints, guides clear-structure depth maps. • Experiments show CrysFormer++ outperforms existing methods in transparent object depth estimation, generates high-precision depth maps under complexity, and offers low-cost perception for robotic manipulation.

Journal

Expert Systems with Applications cover
Expert Systems with Applications
IF:
7.5
Papers:
2.9W
Citations:
10.2W

Organization

No organization information available