Skip to main content

Table 3 Comparison of different loss functions on the PLM validation set

From: Automatic damage identification of Sanskrit palm leaf manuscripts with SegFormer

Method

class IoU (%)

mIoU (%)

INC

BRE

FIB

CON

IMP

CE loss

54.8

34.7

41.4

29.8

91.8

50.5

WCE loss

56.5

41.6

42.6

34.6

93.7

53.8

Focal loss

54.4

39.6

33.9

28.5

93.0

49.9

Dice loss

26.1

8.2

11.1

24.7

16.6

17.3

Combo loss

57.0

42.7

43.4

33.8

93.5

54.1

  1. For comparing loss functions, the resizing and cropping method with a crop size of 512 × 768 was used in all experiments. Best scores are highlighted in bold