Table 3.

Performance measures of MiT models with and without pre-training for data resolutions of 2x, 5x, and 10x.

Data resolution	Model	Pre-trained	F1 score	mIoU	Average accuracy
2x	MiT-B0	False	0.844	0.402	0.518
	MiT-B0	True	0.867	0.609	0.712
	MiT-B2	False	0.782	0.381	0.487
	MiT-B2	True	0.868	0.587	0.746
	MiT-B3	False	0.827	0.371	0.472
	MiT-B3	True	0.869	0.624	0.761
	MiT-B5	False	0.779	0.305	0.409
	MiT-B5	True	0.881	0.639	0.731
5x	MiT-B0	False	0.868	0.428	0.536
	MiT-B0	True	0.899	0.627	0.745
	MiT-B2	False	0.808	0.399	0.500
	MiT-B2	True	0.897	0.604	0.762
	MiT-B3	False	0.840	0.392	0.491
	MiT-B3	True	0.893	0.640	0.772
	MiT-B5	False	0.803	0.327	0.428
	MiT-B5	True	0.907	0.651	0.762
10x	MiT-B0	False	0.873	0.435	0.546
	MiT-B0	True	0.907	0.631	0.754
	MiT-B2	False	0.852	0.404	0.507
	MiT-B2	True	0.909	0.630	0.770
	MiT-B3	False	0.846	0.400	0.498
	MiT-B3	True	0.905	0.642	0.780
	MiT-B5	False	0.812	0.330	0.441
	MiT-B5	True	0.911	0.654	0.774