Multisemantic Level Patch Merger Vision Transformer for Diagnosis of Pneumonia

Table 2

Result comparison between models: precision, recall, and -score center.

Model	ViT	Patch merger	Patch fuser	Label smoothing	Image enhancement	Precision	Recall	-score
M1	X	X	X	X	X	0.886129	0.835897	0.851442
M2	✓	X	X	X	X	0.779524	0.733761	0.744629
M3	✓	✓	X	X	X	0.893820	0.868803	0.878513
MP-ViT	✓	✓	✓	X	X	0.905961	0.875214	0.886710
Final model	✓	✓	✓	✓	✓	0.918173	0.893590	0.903365