Wearing　helmets　has　been　a　mandatory　guideline　for　construction　sites,　mining,　drilling　and　other　fields.　For　different　complex　scenes,　there　is　a　wide　demand　for　efficient　and　accurate　helmet　detection,　and　the　current　helmet　detection　algorithms　have　problems　such　as　large　number　of　parameters　and　weak　timeliness.　In　this　regard,　a　lightweight　Faster-GAMv8　helmet　detection　algorithm　is　proposed　to　improve　YOLOv8.　By　introducing　the　GAM　attention　mechanism,　the　correlation　between　different　channels　is　effectively　captured.　The　method　of　using　FasterBlock　module　to　replace　C2f　significantly　reduces　the　model　parameters　and　FLOPs　operations.　Finally,　the　inclusion　of　the　Wise-IoU　loss　function　enables　a　more　accurate　evaluation　of　the　detection　results　of　small　targets.　Experiments　using　the　VOCdevkit　dataset　show　that　the　proposed　algorithm　achieves　a　mAP　of　91.1%　and　the　floating-point　computation　is　reduced　to　7.6MB,　which　is　able　to　better　satisfy　the　demand　for　fast　detection　at　industrial　sites　and　provides　an　effective　solution　for　real-time　monitoring　in　the　field　of　industrial　safety.　©　2024　IEEE.