CHP-YOLO: An Optimized Deep Learning Methods for Safety Protection Equipment Detection

In the working environment of grain depots, ensuring that personnel safety protection equipment wearing correctly is crucial for safe production. Aiming at the problem of granary personnel wearing detection, this paper uses web crawlers and autonomous photography to construct a SPE (Safety Protection Equipment Detection) dataset. By replacing the YOLOv8 backbone C2f with the C2f_CCNeck module based on CondConv(Conditional Parameterized Convolution), the new YOLO network is lightweight. Then, the improved H-CBAM attention mechanism, and the pruning technology of BatchNormal are integrated. The improved lightweight YOLO detection method CHP-YOLO is proposed. Through ablation experiments, the improved CHP-YOLO shows relatively excellent detection performance. This method effectively improves the accuracy and robustness of Safety Protection Equipment Detection.