Research on Target Detection Algorithm Based on Deep Learning
LI Shuxia
YANG Juncheng
Abstract：This paper improves the generation model generator and discrimination model discriminator of the Generative Ad-versarial Network(eg GAN)network according to the working principle of the GAN network and the characteristics of the Fast R-CNN model.Combining the idea of adding label constraints to Auxiliary Classifier GAN(eg AC-GAN)and Deep Convolutional GAN(eg DC-GAN),and using Convolutional Neural Network(eg CNN)to replace the characteristics of multi-layer perceptron in GAN in both generator and discriminator,this paper proposes Auxiliary Classifier Deep Convolutional GAN(eg AC-DCGAN)mod-el.This model is added with multi-classification and condition auxiliary options and batch normalization operation in the generation model and discrimination model,and uses convolution and deconvolution instead of a pooling layer,and uses global pooling layer instead of a full connection layer.It has been tested on the COCO data set,Pascal VOC 2007 data set,and Pascal VOC 2012 data set,and these tests have achieved good results.
Keywords：generative adversarial networkconvolutional neural networksAC-GANDC-GANAC-DCGAN
Publication Date：2025-10-20
Online Publishing Date：2026-01-16(First online date of this platform, not the publication date of the document)
Pages：5( 2688-2692 )