Lightweight express package detection with G-YOLO feature fusion
ID：162
Submission ID：46 View Protection：ATTENDEE
Updated Time：2025-11-03 11:40:05
Hits：63
Poster Presentation
Start Time：2025-11-09 09:11 （Asia/Shanghai）
Duration：1min
Session：[P] Poster presentation » [P6] 6.AI-driven technology
No files
Abstract
Addressing the challenges posed by low accuracy, a high false detection rate, and the need for lightweight solutions arising from susceptibility to noise and occlusion in complex scenarios, we propose an enhanced method for detecting express packages based on YOLOv5. To strike a more optimal balance between model performance and lightweight design, we introduce the GhostCTRBottleneck module. This module is designed to comprehensively capture feature dependencies while concurrently reducing computational overhead and parameter complexity. Our proposed method demonstrates a 2.5% increase in the MAP index compared to the original YOLOv5s, as demonstrated on a self-built express parcel dataset. Simultaneously, the model's weight, computational load, and parameter count are effectively reduced. Rigorous experiments conducted on the PASCAL VOC 2012 dataset underscore the efficacy and robustness of our method.
Keywords
YOLOv5, Express package detection, Lightweight, GhostCTRBottleneck
Speaker
Comment submit