无锚框的轻量级遥感图像目标检测算法

现有遥感图像目标检测算法存在参数量大、检测速度慢和难以部署于移动设备的问题,为此,本文提出了一种无锚框的轻量级遥感图像目标检测算法.首先设计了DWS-Sandglass轻量化模块以降低模型体积,并改进模型激活函数,以确保检测精度.然后引入无参数注意力模块SimAM,使网络能够专注于更重要的特征信息.最后对无锚框算法的冗余通道进行剪枝操作以减少模型参数量,并通过微调回升精度.在HRSC2016数据集上的实验结果表明,与当前主流的无锚框检测算法相比,该算法在检测精度相当的情况下检测速度更快、模型体积更小,更适合在移动设备部署.

Lightweight remote sensing image target detection without anchor frame

The existing remote sensing image object detection algorithms have been frustrated by large parameter quantities,slow detection speed and inability to deploy on mobile devices.Here,we propose a lightweight remote sensing image object detection algorithm without anchor frames.First,a DWS-Sandglass lightweight module is designed to reduce the model volume,and the activation function of the model is improved to ensure detection accuracy.Then,a parameter free attention module SimAM is introduced to enable the network to focus on more important feature information.Finally,the redundant channels of the anchor frame free algorithm are pruned to reduce the number of model parameters,and the accuracy is improved by fine tuning.The experimental results on HRSC2016 dataset show that compared with current mainstream detection algorithms free of anchor frame,the proposed algorithm has faster detection speed and smaller model size,making it more suitable for deployment on mobile devices with comparable detection accuracy.