Table 1 Evaluation results of the state-of-the-art segmentation networks FCN, U-Net and DeepLab on RGB frames from ESD47.

From: A neuromorphic dataset for tabletop object segmentation in indoor cluttered environment

Datasets

FCN19

U-Net20

DeepLab21

Acc

mIoU

Acc

mIoU

Acc

mIoU

ESD-1 Known obj (ours)

81.37

59.36

86.27

64.19

90.59

68.77

ESD-2 Unknown obj (ours)

64.21

32.79

69.05

40.70

72.16

43.04

MSCOCO17

71.6

31.43

77.2

47.21

79.13

58.01

PascalVoc13

87.09

62.20

92.05

72.70

96.52

87.30

Cityscape18

84.3

65.30

89.07

73.50

93.17

82.10

  1. Furthermore, benchmarks of the same networks on other public datasets MSCOCO, PascalVoc and CityScape are also provided.
Back to article page