Table 8 The comparison of the CR-RNN architectures with respect to original methods in terms of total parameters and different error rates on KITTI 2015 test set
From: Refinement of matching costs for stereo disparities using recurrent neural networks
Architectures | B (%) | EPE | GD(%) | |||
|---|---|---|---|---|---|---|
|  | > 5px | > 4px | > 3px | > 2px |  |  |
MC-CNN-Fast | 11.79 | 12.51 | 13.74 | 16.34 | 5.8471 | 35.46 |
MC-CNN-Acrt | 9.65 | 10.21 | 11.21 | 13.67 | 5.1775 | 35.69 |
MC-CNN-Fast + CR-RNN C | 9.24 | 9.95 | 11.24 | 14.17 | 3.1388 | 27.60 |
MC-CNN-Fast + CR-RNN P | 9.02 | 9.76 | 10.95 | 13.37 | 4.6316 | 22.23 |
SAD | 27.81 | 29.92 | 32.73 | 37.10 | 9.8349 | 62.01 |
SAD + CR-RNN P | 26.72 | 28.42 | 30.77 | 34.71 | 9.6686 | 60.74 |
CBCA | 25.88 | 27.91 | 30.64 | 35.05 | 9.0861 | 56.12 |
CBCA + CR-RNN P | 21.24 | 23.03 | 25.74 | 30.53 | 7.3726 | 45.43 |