Multisemantic Level Patch Merger Vision Transformer for Diagnosis of Pneumonia

Table 1

Result comparison between models: accuracy.

Model	ViT	Patch merger	Patch fuser	Label smoothing	Image enhancement	Accuracy
M1	X	X	X	X	X	0.868590
M2	✓	X	X	X	X	0.777244
M3	✓	✓	X	X	X	0.889423
MP-ViT	✓	✓	✓	X	X	0.897436
Final model	✓	✓	✓	✓	✓	0.911859