Skip to main content

Table 4 The performance of the proposed network with different time step size values on KITTI 2012 test dataset

From: Refinement of matching costs for stereo disparities using recurrent neural networks

Time steps

Input sizes

Param. counts

B (%) > 5px

B (%) > 4px

B (%) > 3px

B (%) > 2px

1

192

164,288

13.53

14.36

15.65

18.66

2

96

188,864

13.29

14.19

15.52

18.47

3

64

229,824

13.33

14.13

15.41

18.52

4

48

274,880

13.21

14.11

15.37

18.19

8

24

465,344

12.95

13.86

15.14

18.01

12

16

659,904

12.89

13.78

15.13

18.26