Table 3.

Performance comparison of different models in KAC.

Name	Accuracy	Recall	Precision	F1-Score	AUC
Binary classification
ResNet18 (10)	0.7834	0.7952	0.7759	0.7854	0.8820
ResNet34 (10)	0.7771	0.7824	0.7639	0.7730	0.8783
DenseNet (32)	0.8122	0.8299	0.8128	0.8212	0.7452
SENet_18 (36)	0.7628	0.7673	0.7478	0.7574	0.8678
ECANet_18 (37)	0.7951	0.7799	0.7648	0.7723	0.8680
HRNet (39)	0.7628	0.7664	0.7543	0.7603	0.8499
GhostNet (35)	0.7538	0.7469	0.7311	0.7389	0.8393
RepVGG (30)	0.8104	0.8149	0.7950	0.8048	0.9158
Vision Transformer (11)	0.7340	0.6943	0.7034	0.6988	0.7465
CONVADD (15)	0.6827	0.6470	0.6527	0.6498	0.6848
Efficient B2 (41)	0.7333	0.6605	0.6846	0.6723	0.7452
Our proposed	0.8431	0.8667	0.8486	0.8575	0.9357
Multi classification
ResNet18 (10)	0.7549	0.7976	0.8186	0.8080	0.9353
ResNet34 (10)	0.7712	0.8220	0.8485	0.8349	0.9589
DenseNet (32)	0.6454	0.6685	0.7289	0.6974	0.8235
SENet_18 (36)	0.5767	0.5940	0.6434	0.6177	0.7666
ECANet_18 (37)	0.6650	0.6895	0.7305	0.7094	0.8785
HRNet (39)	0.7320	0.7658	0.7940	0.7796	0.9226
GhostNet (35)	0.5898	0.6225	0.6913	0.6551	0.8184
RepVGG (30)	0.7173	0.7459	0.7717	0.7586	0.8872
Vision Transformer (11)	0.4771	0.4696	0.5421	0.5034	0.6256
CONVADD (15)	0.7216	0.5926	0.6268	0.6092	0.8339
Efficient B2 (41)	0.6862	0.7207	0.7869	0.7524	0.9114
Our proposed	0.7924	0.8422	0.8596	0.8508	0.9635