Research on Objective Detection Algorithm of Colorectal Lesions
At present,colonoscopy is mainly judged by the naked eye of physicians,which leads to problems such as misdiagnosis and missed diagnosis of the lesions.To solve these problems,an improved algorithm based on YOLOv5s is proposed.Firstly,Silu is used to replace Relu,and the idea of structural reparameterization is added to improve the RepVGG network,and the improved RepVGG network replaces CSPDarknet53 as the Backbone of the algorithm,which effectively inhibits the phenomenon of gradient disappearance and speeds up the model reasoning speed.Finally,in order to improve the sensitivity to small targets,the Head stage of YOLOv5s was improved and added as a four-way detection mechanism.The experimental results show that the improved algorithm can effectively assist physicians to diagnose the disease condition and capture the lesion in time.
万方数据
维普
