Ceramic pixel-level defect detection network based on label decoding enhancement

Yao Chen

doi:10.1016/j.measurement.2026.121500

Ceramic pixel-level defect detection network based on label decoding enhancement

2026-04-13

0

PRE

AI

DOI：10.1016/j.measurement.2026.121500

Original

Original request for help

Save

Abstract

En 中文

• Anovel pixel-level ceramic defect detection network is proposed, integrating label decoding, structural priors, andadaptive feature fusion for fine-grained visual inspection. • Label-Decoding Feature Extraction (LDFE) module leverages state-space modeling and label decoupling toenhance both local detail capture and long-range dependency modeling. • Body Prior Module (BPM) explicitly incorporates morphological priors to improve recognition of overall ceramicstructures and incomplete regions. • Adaptive Feature Interaction (AFI) module dynamically adjusts fusion weights between body and detail branchesto achieve feature co-evolution.

Keywords:

pixel-level defect detection

label decoding

ceramic inspection

feature fusion

structural priors

Ceramic pixel-level defect detection network based on label decoding enhancement

Abstract

Journal

Organization