Table 7 Comparison of different works under the same dataset

[10]

ODIR-5 K

N,D,G,C,AMD,H,M,O

Attention-based unilateral and bilateral feature weighting and fusion network(AUB-Net)

Kappa: 0.640,

F1 Score: 0.913,

AUC value: 0.934

[29]

ODIR-5 K

N,D,G,C,AMD,H,M

ResNet

Accuracy: 0.93,

Sensitivity: 0.84,

Specificity: 0.95,

AUC value: 0.90

[46]

ODIR-5 K

N,D,G,C,AMD,H,M,O

Deep CNN

F1 Score: 0.85,

Kappa score: 0.31,

AUC value: 0.805

[53]

ODIR-5 K

N,C,AMD,M

CNN + 2 Fully Connected Layers

Accuracy: 0.883(95CI (0.812–0.955))

Precision: 0.769(95%CI (0.638–0.901))

Recall: 0.769(95%CI (0.62–0.918))

F1 Score: 0.384(95%CI (0.315–0.454))

CNN + 5 Fully Connected Layers

Accuracy:0.766

Precision: 0.573(95%CI (0.322–0.825))

Recall: 0.542(95%CI (0.361–0.723))

F1 Score: 0.271(95%CI (0.174–0.368))

[31]

ODIR-5 K

N,D,G,C,AMD,H,M,O

DenseNet+multiscale transfer connection (MTC) + domain-specific adversarial adaptation (DSAA)

Accuracy: 0.945(95%CI (0.904–0.985))

AUC value: 0.938(95%CI (0.928–0.949))

F1 Score: 0.929(95%CI (0.917–0.941))

Kappa: 0.697(95%CI (0.663–0.732))

This paper

ODIR-5 K

N,D,G,C,AMD,H,M

ResNet50 + ResNet101 + ID-SET

Accuracy: 0.9237

Precision: 0.945(95% CI (0.92.8–0.963))

Recall: 0.89(95%CI (0.821–0.958))

Specificity: 0.98(95%CI (0.95–1))

AUC value:0.987

F1 Score: 0.914(95%CI (0.875–0.954))

Kappa: 0.878