Table 4 Comparison with state-of-the-art methods on the SAIOD dataset.

Pretrained models

 AlexNet^12,27

85.92

 SqueezeNet^12,28

88.52

 GoogleNet^12,29

89.40

 ResNet-50^12,20

86.40

Transformer-based models

 ViT^12,22

90.00

 Swin-transformer^12,30

90.40

 SwinSight Net (Pradhan et al.¹²)

93.16

 Yolo-based models

 Pradhan et al.¹³

95.33

Proposed models

 Proposed model cross attention

95.52

 Proposed model based MHA

95.80