Table 9 Performance of the GRU-based classifier in signer-independent mode when trained with data from 4 signers and tested with data from 1 signer.

From: A deep learning framework for Ethiopian sign language recognition using skeleton-based representation

Label

Precision

Recall

F1-Score

Support

ሀዘን

0.80

1.00

0.89

35

ህመም

0.17

0.25

0.20

40

ለምን?

0.30

0.94

0.45

36

ልምምድ

0.42

0.35

0.38

37

መራራ

0.79

1.00

0.89

31

መናደድ

0.14

0.03

0.05

35

መጀመር

1.00

0.09

0.17

33

መገረም

0.73

1.00

0.84

38

መጥፎ

0.48

0.30

0.37

37

መጨረስ

0.71

0.65

0.68

37

መጨነቅ

0.36

0.14

0.20

35

ማሸነፍ

0.30

0.38

0.34

37

ምስጋና

0.00

0.00

0.00

38

ቀላል

0.38

1.00

0.55

37

ተነስ!

0.48

0.30

0.37

37

አስቸጋሪ

0.84

0.73

0.78

37

እህት

0.50

0.12

0.20

32

ወንድም

0.38

0.17

0.23

36

ይቅርታ

0.66

1.00

0.80

37

ጎበዝ

0.17

0.06

0.09

35

Accuracy

  

0.48

720

Macro Avg

0.48

0.48

0.42

720

Weighted Avg

0.47

0.48

0.42

720