基于PP-PicoDet技术的智能垃圾分类
2023,31(10):291-299
摘要：目前在垃圾分类目标检测上大多采用YOLOv5系列算法，该算法在相同参数量的情况下，检测精度和检测速度都相对不太高，难以满足实际应用需求。论文对基于PP-PicoDet技术的垃圾分类目标检测应用进行了研究，并将其与几种常见的垃圾分类目标检测算法进行实验对比分析；结果表明，PP-PicoDet算法能够在使用更少的参数量的情况下，实现较高的检测精度和速度，能够满足移动端部署需求。
关键词：垃圾分类；深度学习；目标检测；YOLOv5；PP-PicoDet轻量级模型
Intelligent garbage classification based on PP-PicoDet technology
Abstract：Currently, most of the garbage classification target detection uses the YOLOv5 series algorithm, which has relatively low detection accuracy and detection speed for the same scale parameters and is difficult to meet the practical application requirements. The paper researches the application of garbage classification target detection based on PP-PicoDet technology, and compares it with several algorithms for experimental analysis; the results show that PP-PicoDet algorithm can achieve higher detection accuracy and speed with less number of parameters, and can meet the requirements of deployment on mobile devices.
Key words：garbage classification; deep learning; target detection; YOLOv5; PP-PicoDet
收稿日期：2023-04-14
基金项目：国家自然科学基金项目(12274113)
