notesum.ai

Published at December 5

Exploring Fully Convolutional Networks for the Segmentation of Hyperspectral Imaging Applied to Advanced Driver Assistance Systems

cs.CV

cs.AI

cs.LG

eess.IV

Released Date: December 5, 2024

Authors: Jon Gutiérrez-Zaballa¹, Koldo Basterretxea¹, Javier Echanobe¹, M. Victoria Martínez¹, Inés del Campo¹

Aff.: ¹University of the Basque Country

Arxiv: http://arxiv.org/pdf/2412.03982v1

Refer to caption

U-Net

ANN

Patches

(128x128x25)

Rebuilt images from

overlapping patches

Pixels

(1x1x25)

Accuracy

Precision

IoU

Accuracy

Precision

IoU

Accuracy

Precision

IoU

Road

97.90

95.66

93.74

98.54

94.56

93.25

85.10

92.51

79.62

Road Marks

90.25

73.11

67.75

87.89

77.22

69.80

68.10

21.90

19.86

No Drivable

91.07

97.16

88.71

91.20

98.57

90.01

86.46

89.32

78.42

Overall

95.37

95.55

91.31

95.42

95.44

91.50

85.14

89.32

77.46

Mean

93.07

88.64

83.40

92.54

90.12

84.35

79.89

67.93

59.30

Weighted

90.60

75.71

70.27

88.54

79.43

72.60

70.04

29.36

26.27

Road

92.61

99.05

91.36

93.28

99.00

92.41

73.29

94.30

70.18

Road Marks

80.93

75.39

64.02

78.32

79.11

64.90

68.74

17.50

16.21

Vegetation

94.98

94.63

90.12

95.74

95.80

91.88

93.84

91.47

86.29

Sky

97.86

93.09

91.23

97.49

93.39

91.20

91.04

74.53

69.44

Other

84.97

62.71

56.45

84.83

64.59

57.90

56.18

42.94

32.17

Overall

91.79

93.29

86.47

92.75

93.80

87.66

77.02

85.24

68.45

Mean

90.18

84.98

78.64

89.93

86.38

79.66

76.62

64.15

54.86

Weighted

88.64

82.14

75.27

87.40

84.15

75.93

75.28

44.00

39.55