Authors

Abstract

ABSTRACT To address the issues of large number of parameters and low deployment efficiency on mobile devices in the existing YOLOv8-DSFF model for corn pest detection, this study proposes an improved object detection model that integrates the CSPPC lightweight module and the Wise-IoUv3 loss function. The optimized model reduces the number of parameters by 85.6%, achieves an mAP@0.5 of 90.8%, reaches 204 FPS inference speed on PC and 42 FPS on mobile devices. This provides a practical low-power solution for real-time field monitoring of corn pests.

Abstract in Chinese

摘要 此前玉米虫害识别研究中，YOLOv8-DSFF模型虽较其他检测模型优势显著，但存在参数量大、移动端部署效率低的问题。为此，研究提出融合CSPPC轻量化模块与Wise-IoUv3的目标识别模型改进方案。通过改进模型参数量降低85. 6%；模型mAP@0. 5达90. 8%；PC端推理速度204FPS；移动端帧率42FPS，可为田间实时监测提供低功耗方案。