Skip to main content

Table 3 3D-UNet performance comparison for \(F_{1}\) score, Precision (Pr.), Recall (Re.) with respect to the number of classes

From: Volumetric macromolecule identification in cryo-electron tomograms using capsule networks

 

Validation

Test

 

\(F_{1}\)

Pr.

Re.

\(F_{1}\)

Pr.

Re.

Real data

PT-BG

0.91

0.89

0.92

0.79

0.77

0.81

RB-BG

0.73

0.64

0.85

0.60

0.54

0.67

PT-RB-BG

0.73

0.70

0.77

0.64

0.61

0.68

SHREC’19

4D8Q-BG

0.93

0.89

0.97

0.93

0.90

0.96

1BXN-BG

0.83

0.73

0.95

0.82

0.71

0.98

3GL1-BG\(^{\dagger }\)

0.50

0.39

0.71

0.29

0.22

0.42

4D8Q-1BXN-3GL1-BG

0.80

0.74

0.87

0.69

0.64

0.74

  1. \(^{\dagger }\)Training patch size is \(32 \times 32 \times 32\). More details is in “Case Study” section