A bionic neural network for fish-robot locomotion is presented. The bionic neural network inspired from fish neural net- work consists of one high level controller and one chain of central pattern generators (CPGs)....A bionic neural network for fish-robot locomotion is presented. The bionic neural network inspired from fish neural net- work consists of one high level controller and one chain of central pattern generators (CPGs). Each CPG contains a nonlinear neural Zhang oscillator which shows properties similar to sine-cosine model. Simulation re, suits show that the bionic neural network presents a good performance in controlling the fish-robot to execute various motions such as startup, stop, forward swimming, backward swimming, turn right and turn left.展开更多
针对已有的目标检测方法在复杂场景中对鱼类目标检测效果不理想的问题,提出了一种基于膨胀卷积和参数重构的鱼类目标实时检测方法.先设计了一种四分支融合卷积结构,在引入少量参数量的情况下,扩大了目标检测的感受野,提升了目标检测的效...针对已有的目标检测方法在复杂场景中对鱼类目标检测效果不理想的问题,提出了一种基于膨胀卷积和参数重构的鱼类目标实时检测方法.先设计了一种四分支融合卷积结构,在引入少量参数量的情况下,扩大了目标检测的感受野,提升了目标检测的效果.再引入了RepVGG(重构VGG)并联辅助分支思想,在训练过程中使用复杂模型进行特征学习,而在推理过程中对模型中的BN(Batch Normalization)层以及1×1的辅助分支中的参数进行融合,利用参数重构对训练过程的冗余参量进行合并,保证了模型的低参数量和实时推理.基于YOLOv5s进行实验,相比原始的YOLOv5s获得了更高的检测精度和召回率,平均精度(mean Average Precision,mAP)达到83.1%,超越了目前主流的目标检测算法.提出的算法在检测速度上相比原始模型无明显降低,处理速度上达到100FPS,在实现高精度检测的前提下保证了鱼类目标的实时检测,为基于视觉的鱼类检测方案提供了有效的技术支持.展开更多
文摘A bionic neural network for fish-robot locomotion is presented. The bionic neural network inspired from fish neural net- work consists of one high level controller and one chain of central pattern generators (CPGs). Each CPG contains a nonlinear neural Zhang oscillator which shows properties similar to sine-cosine model. Simulation re, suits show that the bionic neural network presents a good performance in controlling the fish-robot to execute various motions such as startup, stop, forward swimming, backward swimming, turn right and turn left.
文摘针对已有的目标检测方法在复杂场景中对鱼类目标检测效果不理想的问题,提出了一种基于膨胀卷积和参数重构的鱼类目标实时检测方法.先设计了一种四分支融合卷积结构,在引入少量参数量的情况下,扩大了目标检测的感受野,提升了目标检测的效果.再引入了RepVGG(重构VGG)并联辅助分支思想,在训练过程中使用复杂模型进行特征学习,而在推理过程中对模型中的BN(Batch Normalization)层以及1×1的辅助分支中的参数进行融合,利用参数重构对训练过程的冗余参量进行合并,保证了模型的低参数量和实时推理.基于YOLOv5s进行实验,相比原始的YOLOv5s获得了更高的检测精度和召回率,平均精度(mean Average Precision,mAP)达到83.1%,超越了目前主流的目标检测算法.提出的算法在检测速度上相比原始模型无明显降低,处理速度上达到100FPS,在实现高精度检测的前提下保证了鱼类目标的实时检测,为基于视觉的鱼类检测方案提供了有效的技术支持.