Lightweight Underwater Target Detection Algorithm

In order to solve the problem that it is difficult to balance the accuracy and real-time performance of lightweight detection models, a lightweight underwater target detection algorithm was proposed based on the YOLOv8 framework. The Deformable Attention mechanism is introduced to enhance the model's ability to pay attention to the target area, so as to improve the accuracy of target perception. GSConv and VoVGSCSP are used to build an efficient and lightweight neck network, which can achieve effective feature fusion while greatly reducing the computing overhead. The Shared Convolutional Detection Head (LSDH) is designed to effectively reduce the number of parameters and computations while maintaining the detection performance. Experimental results show that the [email protected] of the algorithm in the URPC2020 reaches <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{80.9 \%}$</tex>, which is 1.4 % higher than that of the baseline model, which verifies that the algorithm has both real-time performance and accuracy in underwater detection tasks.