基于ECA-ResNet34的动物种类识别模型

打开文本图片集

中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2025）23-0035-05

Abstract： Aiming at the problems of low recognition accuracy and slow convergence speed of traditional convolutional neuralnetwork inanimalspeciesrecognition，ananimal speciesrecognitionmethodbasedonECA-ResNet34isproposed.This methodusesResNet34asthebasicmodel，integratestheECAatentionmechanismmoduleineachresidualblock，usesonedimensionalconvolution tocapture thedependenciesbetweenchannels，learns theimportance weightsofeachchannel，nd appliesthese weights totheoriginalfeaturemaptoenhancethefeatureresponseof importantchannelsandsuppressthe feature response ofunimportantchannels.TheECA attntionmechanism enhances the attentionofthe model to important feature chanels，andefectivelyimproves theperformanceandgeneralizationabilityofthe model.Theexperimentalresultsshowthat the accuracy of the validation set of the proposed method reaches 94.1% .Compared with the basic model， the accuracy of the validation set is increased by 1.3% ，and the convergence speed is faster and the recognition accuracy is higher.

Keywords：ResNet34;ECA attention mechanism;Deep Learning; image classification

0 引言

我国畜禽养殖业历经人工化、半自动化、自动化到智能化的产业升级进程。（剩余7546字）