Table 5 Comparative results with the SOTA methods on parameters and time–cost on the 3DIRCADb dataset.

From: Deep supervision and atrous inception-based U-Net combining CRF for automatic liver segmentation from CT

Model

Dimension

Parameters

Training time

Test time

Christ et al.3

2D

19,643,746

3 h 23 min 18 s

158.11 s

Chlebus et al.30

2D

7,765,442

2 h 45 min 33 s

139.85 s

Han et al.6

2D

3,858,420

2 h 11 min 04 s

114.22 s

Seo et al.13

2D

4,086,690

3 h 31 min 12 s

137.47 s

Li et al.11

3D

2,059,346,00

42 h 51 min 45 s

45.51 s

Proposed

3D

6,422,037

34 h 01 min 50 s

15.19 s

Proposed + CRF

–

–

34 h 01 min 50 s

2 min 15 s

  1. Significant values are in bold and italics.
Back to article page