Better YOLO with Attention-Augmented Network and Enhanced Generalization Performance for Safety Helmet Detection

TL;DR

The paper tackles safety helmet detection in complex environments by presenting a YOLOv5-based framework that integrates a GhostNetv2 backbone with SCNet and Coordinate Attention modules, plus the Gradient Norm Aware optimizer to enhance generalization. The approach yields a 2% improvement in mAP while reducing parameters and FLOPs by over 25%, demonstrating a favorable accuracy-efficiency balance. Empirical results on Kaggle helmet datasets show robust generalization and effective helmet localization under challenging conditions. This work offers a practical, lightweight solution for real-world safety monitoring in construction and similar settings.

Abstract

Safety helmets play a crucial role in protecting workers from head injuries in construction sites, where potential hazards are prevalent. However, currently, there is no approach that can simultaneously achieve both model accuracy and performance in complex environments. In this study, we utilized a Yolo-based model for safety helmet detection, achieved a 2% improvement in mAP (mean Average Precision) performance while reducing parameters and Flops count by over 25%. YOLO(You Only Look Once) is a widely used, high-performance, lightweight model architecture that is well suited for complex environments. We presents a novel approach by incorporating a lightweight feature extraction network backbone based on GhostNetv2, integrating attention modules such as Spatial Channel-wise Attention Net(SCNet) and Coordination Attention Net(CANet), and adopting the Gradient Norm Aware optimizer (GAM) for improved generalization ability. In safety-critical environments, the accurate detection and speed of safety helmets plays a pivotal role in preventing occupational hazards and ensuring compliance with safety protocols. This work addresses the pressing need for robust and efficient helmet detection methods, offering a comprehensive framework that not only enhances accuracy but also improves the adaptability of detection models to real-world conditions. Our experimental results underscore the synergistic effects of GhostNetv2, attention modules, and the GAM optimizer, presenting a compelling solution for safety helmet detection that achieves superior performance in terms of accuracy, generalization, and efficiency.

Figures (6)

• Figure 1: This fig compares our improved framework with the original YOLOv5 framework. In the backbone section, we have replaced the conv and C3 modules with GhostConv and GhostC3, respectively. Additionally, we have substituted the original SPPF with an SPPF integrated with SCNet. In the neck section, we have added a Coordination and Attention (CA) module following each concatenation step.
• Figure 2: The framework of the GhostNetV2 Bottleneck
• Figure 3: The workflow of SCNet
• Figure 4: The schematic diagram of the CA module
• Figure 5: This figure displays the feature maps generated by different experimental networks for the same test image. These four rows experimental images are respectively from experiments 2, 6, 7, and 8 in Table 1.