基于CycleGAN以及改进YOLOv5的垃圾检测算法OA
Garbage Detection Algorithm Based on CycleGAN and Improved YOLOv5
垃圾分类自2019年以来就是一个热门话题,通过人工智能算法对生活垃圾进行合理分类具有重大意义.人工智能发展至今,从传统的机器学习到现在的深度学习,不论是识别速度还是识别效率都有了大幅提升.为了解决夜晚数据集不足的问题以及提高垃圾检测的识别率,基于CycleGAN进行了数据增强,并对YOLOv5模型进行了改进.实验结果表明,使用CycleGAN数据增强后,mAP由0.795增长到了0.864,增长了6.9%,改进YOLOv5后mAP由0.864增长到了0.903,增长了3.9%.
Garbage classification has been a hot topic since 2019,and it is of great significance to reasonably classify domes-tic garbage through artificial intelligence algorithms.Since the development of artificial intelligence,from traditional machine learn-ing to the current deep learning,both the recognition speed and recognition efficiency have been greatly improved.In order to solve the problem of insufficient datasets at night and improve the recognition rate of garbage detection,data enhancement is carried out based on CycleGAN,and the YOLOv5 model is improved.The experimental results show that after using CycleGAN data enhance-ment,mAP increases from 0.795 to 0.864,an increase of 6.9%,and after improving YOLOv5,mAP increases from 0.864 to 0.903,achieving an increase of 3.9%.
赵和月;何利文
南京邮电大学物联网学院 南京 210003南京邮电大学物联网学院 南京 210003
信息技术与安全科学
垃圾分类深度学习YOLOv5CycleGAN数据增强目标检测
garbage classificationdeep learningYOLOv5CycleGANdata augmentationtarget detection
《计算机与数字工程》 2026 (6)
1575-1579,1624,6
评论