激光与光电子学进展, 2019, 56 (22): 221003, 网络出版: 2019-11-02  

基于反残差结构的轻量级多目标检测网络 下载: 1082次

Light-Weight Multi-Object Detection Network Based on Inverted Residual Structure
作者单位
辽宁工程技术大学软件学院, 辽宁 葫芦岛 125105
摘要
针对YOLO系列的目标检测方法参数多、计算量大、生成检测模型规模大等导致对运行硬件平台计算资源要求高的问题,提出一种基于反残差结构的轻量级多目标检测网络(IR-YOLO)。首先,利用深度可分离卷积减少模型参数和计算量;其次,基于深度可分离卷积构造反残差模块,提取高维特征;最后,根据反残差结构特点,利用线性激活函数减少通道组合过程激活函数的信息损失。IR-YOLO算法较YOLOv3-Tiny算法模型尺寸减少47.7%。实验结果表明IR-YOLO算法在不影响检测精度的前提下,可有效减少网络计算量和存储量。
Abstract
To solve high computational resource requirement for running hardware platform of the series of the YOLO object detection method due to the huge parameters, the large amount of calculation, and the large scale of detection model, this paper developed a light-weight object detection network based on inverted residual structure(IR-YOLO). First, it used depth separable convolution to reduce detection model parameters and computational quantities. Secondly, it constructed inverted residual block based on depth separable convolution to extract high-dimensional feature. Finally, according to the characteristic of inverted residual structure, it used a linear activation function to reduce the information loss during the process of channels combination. The experimental results show that the IR-YOLO detection model is reduced by 47.7% compared to the YOLOv3-Tiny detection model, it validated that the IR-YOLO algorithm can effectively compress the model while maintaining detection accuracy.

刘万军, 高明月, 曲海成, 刘腊梅. 基于反残差结构的轻量级多目标检测网络[J]. 激光与光电子学进展, 2019, 56(22): 221003. Wanjun Liu, Mingyue Gao, Haicheng Qu, Lamei Liu. Light-Weight Multi-Object Detection Network Based on Inverted Residual Structure[J]. Laser & Optoelectronics Progress, 2019, 56(22): 221003.

引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!