一种改进的DSOD目标检测算法

An Improved DSOD Object Detection Algorithm

针对DSOD目标检测算法对小目标检测能力较弱的问题,提出在DSOD中引入RFB_a网络模块和Atrous卷积层予以改进。首先,该算法将DSOD网络的第二个转接层产生的特征图输入到RFB_a网络模块中,经过RFB_a网络不同采样步长的Atrous卷积提取具有不同感受野的特征,为后续检测小目标步骤提供所需特征;其次,为了增加特征图的语义信息,在第二个无池化转接层后加入采样步长为6的Atrous卷积层;最后,在损失函数中加入IOG惩罚项,防止在预测密集的同类型目标时出现同类预测框重叠,从而避免在NMS后处理时出现漏检。实验表明,该算法相对于原DSOD算法具有更高的检测精度,提高了对小目标的检测能力,同时降低了训练网络的硬件设备要求。

In order to improve the detection performance of DSOD for small objects,the RFB_a network module and Atrous convolution were introduced into DSOD.Firstly,the improved algorithm inputs the feature map generated by the second transition layer of the DSOD network into the RFB_a network module,and extracts the features with different receptive fields through the Atrous convolution with different sampling steps of the RFB_a network for small objects detection.Then,in order to improve the semantic information in the feature map,an Atrous convolution layer with rate size of 6 was added after the second transition layer without pooling.Finally,an IOG penalty term was added to the loss function to prevent the prediction box from overlapping when predicting objects of the same class,which may leads to detection missing after NMS processing.Experimental results show that,compared with the original DSOD algorithm,the improved algorithm can obtain higher detection accuracy and better detection ability of small objects,and meanwhile reduce the hardware requirements on the training network.