摘要
随着深度学习在计算机领域应用层次的不断加深,在图像分类、目标检测和跟踪等领域所使用的卷积神经网络的深度也随之加大。轻量级网络的提出很大程度上解决了网络模型过大的问题,已被广泛应用到图像分类、目标检测等领域。文章设计一种新型无padding的暹罗结构轻量级网络框架,融合SiamFC目标跟踪框架,模型大小缩减为原算法的三分之一,精度和成功率分别提高0.34、0.12,跟踪速度达到102帧/秒。
With the continuous deepening of the application level of deep learning in the computer field,the depth of convolutional neural networks used in image classification,object detection,and tracking has also increased.The proposal of lightweight networks has largely solved the problem of excessively large network models and it has been widely applied in fields such as image classification and object detection.This paper designs a new type of non padding Siamese lightweight network framework that integrates the SiamFC target tracking framework.The model size is reduced to one-third of the original algorithm,and the accuracy and success rate are improved by 0.34 and 0.12,respectively,the tracking speed reaches 102 FPS.
作者
徐文豪
张秀梅
王振兴
XU Wenhao;ZHANG Xiumei;WANG Zhenxing(School of Physics and Electronic Information,Dezhou University,Dezhou 253023,China)
出处
《现代信息科技》
2023年第17期84-87,共4页
Modern Information Technology
关键词
轻量级网络
暹罗网络
目标跟踪
lightweight network
Siamese network
target tracking
作者简介
徐文豪(1996-),女,汉族,山东临沂人,助教,硕士,研究方向:目标检测与跟踪。
