EURASIP Journal on Image and Video Processing

Table 9 The comparison of the CR-RNN architectures with respect to original methods in terms of and error rates (%) on KITTI 2012 and KITTI 2015 online test set

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

	KITTI 2012				KITTI 2015
Architectures	> 5px	> 4px	> 3px	> 2px	D1-bg	D1-fg	D1-all
MC-CNN-Fast	15.77	16.52	17.58	19.72	15.38	28.84	17.62
MC-CNN-Acrt	14.17	14.77	15.65	17.62	13.29	28.29	15.79
MC-CNN-Fast + CR-RNN _P	13.33	13.83	14.54	15.99	12.98	28.36	15.54
CBCA	31.00	32.52	34.51	37.95	30.81	35.51	31.59
CBCA + CR-RNN _P	26.58	27.97	29.98	33.92	26.56	30.71	27.25

Back to article page