Table 9.

Average performance metrics of the cutting edge approaches for the thermal image in detecting DG subjects

Algorithms	Performance metrics
	Accuracy	Precision	Recall	F1-score	Computational cost (sec/image)
TransUNet	0.86	0.78	0.89	0.83	0.46
Swin-UNETR	0.83	0.79	0.752	0.782	0.50
MedT (Gated Axial Transformer)	0.872	0.87	0.767	0.77	0.47
DE-ResUNet	0.80	0.79	0.79	0.775	0.57
Hybrid CNN-ViT (ResNet-ViT)	0.89	0.841	0.85	0.894	0.49
ViT-Caps	0.975	0.94	0.87	0.85	0.42
Proposed Model	0.987	0.97	0.95	0.952	0.37