Tang, Yaocheng (Minzu University of China), Zhang, Xiaohan (Zhejiang University), Hu, Xiaoxi (Beijing Jiaotong University), Zheng, Yutong (School of Information Engineering, Minzu University of China, Be)

CADet: Crack-Aware Detection with Dual-Decoder and Probability-Distribution-Based Fusion

Scheduled for presentation during the Video Session "On-Demand Video Presentations" (VP-VP), Saturday, November 22, 2025, 08:00−18:00, On-Demand Platform

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on April 2, 2026

Abstract

Most previous crack detection networks typically use a receptive-field-expansion module for global context integration and a feature-fusion module for local detail refinement. However, we find that the architecture which combines these two modules within a single decoder limits the crack detection performance. To address this, we propose a Crack-Aware Detection Network (CADet), which decouples receptive field expansion and feature fusion through a dual-decoder architecture: one branch dedicated to global context aggregation (receptive field expansion) and the other optimized for local detail refinement (feature fusion). To effectively fuse outputs from the dual-decoder, we further propose a Probability-Distribution-Based Fusion strategy that fuses the dual-decoder outputs through probabilistic mapping to generate the prediction result. Extensive experiments on four public datasets demonstrate that CADet achieves state-of-the-art performance. Code is Available at https://github.com/YiZhiTangZong/CADet.