Lightweight Underwater Target Detection for Channel Spatial Depth Perception
A lightweight underwater target detection network CSDP-L-YOLO for channel spatial depth perception is proposed.The network is improved based on the YOLOv5 network and consists of the feature awareness module and two-attention gating strate-gy.The feature sensing module aims to adaptively suppress or enhance multi-level features in the decoder,optimize the consistency of in-class learning,and solve the false detection and missing detection caused by the complexity of underwater scenes.The feature map-ping is generated by the linear operation and mixing structure to reduce the fusion and calculation of redundant features,so as to re-duce the parameters and computational complexity of the model.The dual attention gating strategy introduces both the concurrent channel space squeezing excitation mechanism module and convolutional attention module into the encoder simultaneously,so as to further focus on the strong correlation features and enhance the sensitivity of the model on the features.Experimental results show that compared with the baseline model,the proposed model improves the mean average precision(mAP)by 2.4％,saves the parame-ters by 20％and the computation by 15.8％,and improves the detection speed by 8.2 ms.In addition,compared to currently more advanced YOLOv8 model,the mAP of the proposed model improves by 1.9％.
万方数据
维普
