Table 1 LodeSTAR has discriminative power

From: Single-shot self-supervised object detection in microscopy

Model particle

Point

Sphere

Annulus

Ellipse

Crescent moon

Point

0.02

31.44

70.47

12.82

1590.37

Sphere

857.96

0.05

58.66

1.13

1025.23

Annulus

58.60

26.94

0.07

27.45

750.94

Ellipse

758.71

0.85

75.16

0.12

21.77

Crescent

13.97

11.47

18.32

10.10

0.18

  1. Each model trained by LodeSTAR is significantly more self-consistent (lower average weighted variance) when presented data from its training distribution (bold font), compared to when it is evaluated on different data. This demonstrates that LodeSTAR has acquired discriminative power.
Back to article page