Hard exudate (HE) is of great importan ce for the early diagnosis of diabetic retinopathy (DR),and the accurate detecti on of HE from the fundus image is an important step in the diagnosis and screeni ng of DR.To solve the problems of poor segmentation effect and high false detect ion rate of small targets in previous automatic detection methods,based on the c lassic U-Net generation network,this paper introduces the chain fusion mechanis m of attention residuals,and constructs a chain fusion of attention residuals ge neration countermeasures network (CFAR-GAN) suitable for retinal HE detection.A network in the generated code after each module of the process of adding a resi dual network (ResNet) structure,and uses the convolution with residual connectio n layer chain (Res path) jump between different layers of connections,at the sam e time the global biggest pooling attention mechanism is used for depicting the weight of different depth characteristics,in order to prevent the training proce ss of the fitting so as to improve the generalization ability of the network.CFA R-GAN was trained with the training set of e-ophtha EX database.In the test se t,the detection sensitivity,PPV and F-score were 92.5%,88.7% and 90.6%,respecti vely.The trained network was tested on another independent DIARETDB1database,an d the sensitivity,specificity and accuracy were 100%,98.5% and 99.1%,respectivel y,indicating that the method proposed in this paper has ideal generalization abi lity,which is of positive significance for accurate and efficient detection of H E in fundus images and early automatic diagnosis of DR.