Table 5 Performance of models using baseline view.

From: A benchmark for domain adaptation and generalization in smartphone-based human activity recognition

Model

Time

Frequency

KH

MS

RW-T

RW-W

UCI

WDM

Mean

KH

MS

RW-T

RW-W

UCI

WDM

Mean

KNN

42.4%

81.0%

31.0%

46.1%

54.6%

60.4%

52.6%

86.8%

88.9%

73.6%

68.9%

81.7%

93.5%

82.2%

Random Forest

80.6%

87.9%

74.1%

77.4%

85.5%

92.1%

82.9%

79.7%

91.0%

83.8%

76.2%

92.2%

97.0%

86.7%

SVM

57.6%

76.9%

81.1%

68.8%

85.7%

92.5%

77.1%

70.1%

84.7%

85.5%

79.9%

86.2%

98.6%

84.2%

CNN (1D)12

77.5%

93.4%

80.8%

73.0%

95.2%

95.6%

85.9%

75.0%

91.2%

82.4%

82.9%

94.5%

96.6%

87.1%

CNN (2D)12

76.0%

93.2%

71.3%

77.7%

95.3%

91.5%

84.1%

79.6%

91.8%

79.6%

78.6%

89.1%

95.5%

85.7%

CNN PF34

79.4%

93.2%

70.7%

73.8%

94.2%

88.1%

83.3%

82.1%

92.1%

77.5%

83.2%

92.9%

96.6%

87.4%

CNN PFF34

79.4%

93.5%

73.4%

72.6%

95.8%

89.7%

84.1%

85.0%

90.4%

78.5%

83.3%

93.2%

96.3%

87.8%

ConvNet13

75.0%

93.5%

68.3%

74.2%

91.9%

90.5%

82.2%

87.6%

91.5%

86.3%

82.3%

95.0%

96.9%

89.9%

IMU CNN14

75.0%

87.4%

60.0%

64.3%

89.6%

84.8%

76.8%

84.2%

91.7%

75.6%

78.7%

94.1%

96.6%

86.8%

IMU Transf.14

74.9%

70.5%

73.0%

74.2%

92.2%

89.4%

79.0%

72.2%

73.1%

78.5%

76.9%

78.8%

96.3%

79.3%

MLP (2 Layers)

75.0%

83.2%

77.8%

63.3%

79.9%

91.4%

78.4%

86.7%

92.7%

82.5%

77.5%

92.8%

97.9%

88.4%

MLP (3 layers)

78.8%

82.7%

76.4%

64.5%

80.8%

88.6%

78.6%

86.2%

90.5%

81.7%

76.9%

93.7%

98.5%

87.9%

ResNet15

79.6%

86.8%

74.6%

76.9%

97.6%

91.9%

84.6%

70.4%

86.0%

80.5%

71.4%

92.8%

93.6%

82.5%

ResNetSE67

78.2%

90.9%

72.2%

76.1%

97.4%

92.9%

84.6%

76.2%

82.7%

80.3%

76.1%

91.6%

94.1%

83.5%

ResNetSE-567

78.6%

89.0%

70.0%

75.3%

95.3%

90.4%

83.1%

76.1%

90.3%

79.0%

78.2%

92.5%

94.5%

85.1%

Max

80.6%

93.5%

81.1%

77.7%

97.6%

95.6%

85.9%

87.6%

92.7%

86.3%

83.3%

95.0%

98.6%

89.9%

  1. The best results for each dataset and for each domain (time and frequency) are highlighted in bold. Mean column represents the average performance of the model in the datasets.