Low-res two-view results

This table lists the benchmark results for the low-res two-view scenario. This benchmark evaluates the Middlebury stereo metrics (for all metrics, smaller is better):

bad 0.5, 1.0, 2.0, 4.0 [%]: Fraction of pixels with errors larger than the given number of disparities.
Average error [px]: The per-pixel average disparity error.
Root mean square error [px]: The per-pixel root mean square disparity error.
50%, 90%, 95%, 99% error quantile [px]: The highest disparity error within the given percentage of best pixels (for 50%, this is the median error).
Time [s]: The runtime of the method.

The mask determines whether the metric is evaluated for all pixels with ground truth, or only for pixels which are visible in both images (non-occluded).
The coverage selector allows to limit the table to results for all pixels (dense), or a given minimum fraction of pixels.

Methods with suffix _ROB may participate in the Robust Vision Challenge.

Click one or more dataset result cells or column headers to show visualizations. Most visualizations are only available for training datasets. The visualizations may not work with mobile browsers.

Method	all	lakes. 1l	lakes. 1s	sand box 1l	sand box 1s	stora. room 1l	stora. room 1s	stora. room 2l	stora. room 2s	stora. room 2 1l	stora. room 2 1s	stora. room 2 2l	stora. room 2 2s	stora. room 3l	stora. room 3s	tunnel 1l	tunnel 1s	tunnel 2l	tunnel 2s	tunnel 3l	tunnel 3s

DN-CSS_ROB	8.17 1	2.59 9	14.14 17	6.11 7	3.22 12	6.31 1	3.53 3	22.92 3	20.59 3	16.66 3	30.05 6	9.31 1	7.95 2	13.49 1	4.78 1	0.72 19	0.00 1	0.45 15	0.00 1	0.54 8	0.09 1

MLCV	9.32 2	2.22 7	12.55 12	5.22 2	1.32 1	13.79 4	1.35 1	21.14 1	23.65 4	21.91 7	31.37 9	17.14 3	7.57 1	16.00 4	10.18 6	0.22 6	0.01 2	0.03 5	0.02 7	0.40 3	0.29 4

iResNet	10.26 3	3.10 16	15.72 24	7.35 13	2.13 4	13.86 5	7.07 11	22.80 2	28.01 6	20.10 5	30.40 7	16.79 2	11.06 3	14.46 3	10.29 9	0.39 12	0.05 7	0.00 1	0.07 11	0.73 10	0.73 11

DeepPruner_ROB	10.77 4	4.56 23	13.13 13	6.37 9	4.28 17	10.14 2	6.86 9	36.42 25	19.59 2	17.87 4	27.37 4	23.19 7	12.26 4	19.09 11	10.23 8	1.05 25	0.48 26	0.23 10	0.15 14	0.89 14	1.17 18

NOSS_ROB	10.99 5	3.81 19	6.88 3	6.69 11	2.30 6	16.06 9	10.94 16	30.06 11	32.41 13	16.10 2	18.72 1	22.78 5	14.21 6	17.52 7	8.62 4	2.35 32	1.84 34	3.32 36	1.57 32	2.06 26	1.52 23

HSM	11.33 6	1.80 2	7.04 5	6.13 8	3.99 15	16.16 10	6.92 10	29.69 9	19.39 1	22.95 8	26.05 2	41.08 25	15.79 10	20.25 14	8.89 5	0.02 1	0.03 5	0.01 4	0.00 1	0.15 1	0.34 6

iResNet_ROB	11.71 7	1.96 5	10.77 9	6.01 6	3.68 13	15.13 8	2.34 2	31.72 13	37.58 22	26.16 12	32.77 13	25.88 11	15.34 8	16.35 5	7.94 3	0.19 4	0.01 2	0.00 1	0.00 1	0.31 2	0.12 2

CBMV_ROB	11.77 8	2.89 11	6.92 4	5.20 1	2.88 9	14.12 6	4.79 5	26.69 4	29.84 8	26.17 13	31.92 11	23.61 8	20.74 22	20.20 13	10.18 6	1.78 30	1.80 33	2.17 33	1.26 30	1.61 21	0.55 9

iResNetv2_ROB	12.00 9	2.91 12	13.46 14	5.82 4	3.73 14	13.15 3	6.45 8	34.15 18	36.02 18	25.25 11	33.81 16	29.79 14	14.24 7	13.61 2	6.10 2	0.21 5	0.02 4	0.12 8	0.00 1	0.82 12	0.25 3

SGM-Forest	12.92 10	1.87 3	6.61 1	5.68 3	2.05 3	23.19 17	11.30 18	34.24 19	30.77 11	26.54 14	31.99 12	29.98 15	15.74 9	21.17 16	13.36 10	1.05 25	0.33 19	0.84 23	0.01 6	0.71 9	0.97 15
Johannes L. Schönberger, Sudipta Sinha, Marc Pollefeys: Learning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching. ECCV 2018
DLCB_ROB	13.35 11	3.00 14	9.12 7	9.43 21	5.68 24	21.80 14	10.12 14	29.19 8	29.92 9	27.71 16	31.45 10	32.37 17	19.50 19	19.02 10	16.73 15	0.22 6	0.04 6	0.38 14	0.06 10	0.49 6	0.85 12

StereoDRNet-Refined	13.36 12	2.80 10	10.82 10	7.94 18	3.10 10	18.95 11	4.45 4	28.47 7	29.65 7	29.11 18	41.47 30	24.44 10	17.64 15	23.34 17	21.66 18	0.16 2	0.07 8	0.53 18	0.32 18	0.77 11	1.64 24
Rohan Chabra, Julian Straub, Chris Sweeney, Richard Newcombe, Henry Fuchs: StereoDRNet. CVPR
CBMV	13.69 13	3.63 17	8.02 6	5.93 5	2.69 7	22.57 15	12.44 19	29.81 10	31.57 12	31.20 25	33.79 15	31.04 16	17.41 14	25.28 21	14.11 11	0.71 18	0.60 28	0.60 19	0.11 12	1.11 17	1.13 17
Konstantinos Batsos, Changjiang Cai, Philippos Mordohai: CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation. Computer Vision and Pattern Recognition (CVPR) 2018
pmcnn	14.59 14	3.68 18	19.78 32	6.83 12	4.51 18	21.27 13	14.67 23	27.37 5	30.54 10	27.09 15	40.19 27	38.71 23	18.53 18	17.70 8	19.27 16	0.84 20	0.09 11	0.00 1	0.00 1	0.48 4	0.31 5

NVstereo2D	15.15 15	2.97 13	16.28 28	9.23 20	7.56 33	30.03 27	9.53 13	42.80 37	42.72 27	15.11 1	27.03 3	23.16 6	16.93 12	24.28 18	15.45 12	3.07 35	0.43 23	1.44 29	0.61 24	6.74 36	7.60 39

StereoDRNet	15.30 16	4.47 22	14.35 20	10.71 26	8.69 36	25.02 21	11.03 17	39.66 30	35.62 17	29.17 19	28.81 5	34.05 19	17.80 16	18.80 9	23.77 24	0.41 13	0.30 18	0.15 9	0.14 13	1.64 22	1.42 21

DRN-Test	15.88 17	3.83 20	14.70 21	9.96 23	8.02 35	28.93 26	14.65 22	42.34 33	38.25 24	28.18 17	34.31 19	29.22 13	18.35 17	20.01 12	22.68 20	0.37 9	0.41 21	0.46 16	0.28 17	1.28 18	1.34 20

DISCO	16.36 18	1.88 4	16.15 26	7.78 17	4.19 16	26.92 23	10.21 15	30.31 12	44.26 31	21.35 6	33.91 17	35.80 22	22.40 23	35.65 38	33.87 39	0.18 3	0.08 9	0.04 6	0.04 8	1.71 23	0.54 8

NaN_ROB	16.51 19	6.51 27	16.22 27	10.53 25	4.64 20	31.76 30	17.78 31	37.00 26	43.37 29	29.68 22	31.29 8	32.67 18	20.32 21	28.00 26	16.06 13	0.38 10	0.41 21	0.32 12	0.41 20	0.87 13	1.94 25

PSMNet_ROB	16.90 20	5.45 25	14.16 18	13.77 32	7.25 32	27.65 24	32.74 41	43.21 38	37.88 23	24.00 9	32.97 14	34.55 20	16.69 11	17.41 6	23.96 26	0.38 10	0.22 14	0.93 25	2.05 35	1.83 24	0.96 14

NCCL2	17.23 21	6.37 26	14.30 19	23.45 42	7.18 31	24.16 19	17.05 29	34.39 21	25.88 5	31.55 26	38.01 21	42.09 27	26.81 29	20.94 15	22.69 21	0.45 14	0.23 15	2.63 34	1.80 33	2.02 25	2.57 30

XPNet_ROB	17.33 22	4.84 24	15.44 23	11.15 27	6.64 28	20.88 12	16.10 27	36.26 24	39.54 25	31.80 27	41.32 29	41.30 26	23.91 24	24.67 19	27.40 33	1.07 27	0.74 29	0.79 22	0.25 15	1.10 16	1.31 19

ETE_ROB	17.67 23	9.48 34	17.66 29	13.60 30	4.57 19	23.03 16	21.09 35	32.49 16	35.24 16	30.49 23	38.06 22	46.18 36	24.49 25	26.40 23	24.21 27	0.46 15	0.24 16	1.14 28	0.84 27	1.52 20	2.13 28

PWCDC_ROB	17.80 24	9.62 35	18.38 30	17.95 36	7.74 34	24.61 20	5.56 7	34.41 22	52.79 40	29.66 21	50.20 36	20.63 4	19.69 20	29.85 28	20.87 17	6.18 37	0.26 17	0.10 7	0.05 9	5.36 35	2.00 26

MDST_ROB	17.92 25	1.60 1	13.68 15	13.88 33	6.44 26	43.05 41	14.86 24	42.74 36	41.66 26	43.25 36	42.85 33	28.43 12	17.28 13	29.35 27	16.06 13	0.88 21	0.11 12	0.63 20	0.47 22	0.50 7	0.64 10

PWC_ROB	18.37 26	10.55 38	25.19 35	10.21 24	6.29 25	23.75 18	4.79 5	35.93 23	48.75 34	32.56 28	44.46 35	34.68 21	24.78 27	30.99 29	25.72 31	1.38 28	0.08 9	1.57 30	0.26 16	2.18 27	3.17 33

Anonymous Stereo	18.60 27	11.75 39	49.81 43	14.43 34	12.02 38	14.33 7	23.23 38	32.16 15	43.08 28	24.32 10	34.16 18	24.24 9	14.12 5	31.67 30	30.84 38	0.89 22	0.91 31	1.74 31	1.92 34	3.23 31	3.23 34

LALA_ROB	19.19 28	7.15 31	15.93 25	12.96 29	5.30 23	28.30 25	23.21 37	42.51 34	36.69 19	33.45 30	40.81 28	50.96 39	24.53 26	27.92 25	25.66 30	0.60 16	0.44 24	2.04 32	1.22 29	2.54 28	1.47 22

SGM_ROB	19.52 29	2.12 6	6.79 2	6.59 10	1.33 2	38.20 40	15.74 26	38.13 28	34.07 15	45.71 38	41.91 31	51.41 40	38.10 38	38.77 40	27.80 35	0.61 17	0.36 20	0.51 17	0.33 19	1.03 15	0.90 13
Heiko Hirschmueller: Stereo processing by semiglobal matching and mutual information. TPAMI 2008, Volume 30(2), pp. 328-341
WCMA_ROB	20.02 30	4.30 21	19.72 31	9.47 22	7.00 30	32.71 35	13.99 21	31.97 14	32.48 14	41.51 35	52.00 38	44.09 30	36.14 37	32.43 32	24.29 28	6.19 38	2.79 38	1.09 27	1.10 28	3.95 32	3.13 32

SANet	20.96 31	6.60 28	29.81 37	8.83 19	3.19 11	31.27 29	20.56 34	41.86 32	56.09 44	39.30 33	43.62 34	44.95 32	31.93 33	27.83 24	23.67 23	0.94 23	0.52 27	0.99 26	0.42 21	4.72 34	2.03 27

FBW_ROB	21.00 32	10.23 37	22.72 33	13.71 31	6.77 29	30.49 28	16.46 28	44.58 39	43.57 30	44.25 37	39.59 25	43.27 29	26.12 28	33.96 34	21.80 19	2.58 34	1.84 34	7.14 39	2.78 37	4.00 33	4.06 35

CSAN	21.34 33	9.45 33	23.34 34	20.99 39	4.95 22	32.57 33	34.26 42	38.83 29	49.95 38	36.97 32	39.72 26	44.97 33	31.73 32	26.05 22	23.94 25	1.52 29	0.47 25	0.85 24	1.43 31	2.73 29	2.14 29

MeshStereo	22.27 34	6.87 29	11.15 11	7.69 15	4.87 21	37.70 39	13.06 20	41.64 31	36.95 20	50.92 41	53.41 39	58.12 41	41.93 40	37.73 39	27.62 34	2.52 33	2.37 37	2.99 35	2.07 36	3.09 30	2.60 31
C. Zhang, Z. Li, Y. Cheng, R. Cai, H. Chao, Y. Rui: MeshStereo: A Global Stereo Model with Mesh Alignment Regularization for View Interpolation. ICCV 2015
MSMD_ROB	22.90 35	9.64 36	14.76 22	17.46 35	9.88 37	34.94 37	18.83 32	33.67 17	37.47 21	40.97 34	59.80 42	45.93 34	38.34 39	31.74 31	23.24 22	6.22 39	3.58 40	8.78 40	11.03 42	6.76 37	5.06 37

LE_ROB	22.93 36	3.07 15	14.02 16	7.70 16	2.86 8	31.99 32	17.35 30	28.10 6	67.19 48	70.51 48	55.61 41	49.07 38	47.62 42	24.76 20	36.49 41	0.35 8	0.20 13	0.27 11	0.55 23	0.48 4	0.44 7

SGM-ForestM	23.72 37	2.44 8	9.13 8	7.43 14	2.25 5	44.80 42	19.11 33	44.90 40	49.97 39	50.74 40	51.18 37	62.06 46	45.83 41	44.96 42	33.88 40	1.00 24	0.84 30	0.78 21	0.62 25	1.29 19	1.11 16

DPSNet	24.40 38	6.97 30	33.14 38	11.16 28	6.54 27	53.33 44	43.32 46	51.28 45	59.37 46	30.89 24	39.36 23	40.35 24	34.30 35	32.57 33	25.09 29	4.65 36	1.62 32	0.37 13	0.66 26	8.51 38	4.44 36

PDISCO_ROB	24.45 39	9.24 32	29.28 36	28.68 43	19.96 41	37.14 38	15.33 25	45.04 41	54.69 42	29.24 20	42.67 32	46.07 35	28.17 30	35.35 36	30.30 37	9.92 40	2.13 36	6.90 38	3.20 38	8.74 39	6.87 38

PASM	28.19 40	17.35 40	48.18 42	22.49 41	22.20 42	25.92 22	25.84 39	34.35 20	49.45 35	35.37 31	39.48 24	46.72 37	33.36 34	34.28 35	36.77 42	11.73 43	14.94 44	17.43 43	22.18 44	13.45 41	12.32 42

LSM	29.98 41	19.52 43	55.11 44	19.06 37	53.75 47	32.57 33	23.17 36	42.61 35	48.57 33	45.97 39	55.04 40	42.93 28	36.02 36	35.63 37	27.03 32	2.33 31	9.19 43	4.87 37	7.71 41	10.65 40	27.91 45

DispFullNet	30.27 42	31.74 44	58.63 45	30.13 44	28.18 44	31.79 31	8.94 12	37.16 27	54.68 41	32.60 29	37.99 20	44.62 31	28.80 31	41.55 41	27.93 36	11.94 44	3.46 39	35.53 45	12.34 43	30.08 45	17.26 43

ELAS	33.68 43	18.95 41	36.71 40	20.64 38	13.61 39	56.43 45	35.14 43	50.03 44	49.90 36	60.58 42	63.83 46	58.64 43	53.36 43	50.73 44	44.44 44	11.03 41	6.58 41	10.02 41	7.43 40	13.58 43	11.94 41
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
ELAS_ROB	33.79 44	19.15 42	35.97 39	21.28 40	13.79 40	52.58 43	36.11 44	48.96 43	55.66 43	61.47 45	62.91 44	58.35 42	53.55 44	53.37 45	42.23 43	11.25 42	6.64 42	10.32 42	7.05 39	13.56 42	11.58 40
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
NVStereoNet_ROB	41.93 45	32.40 45	44.55 41	36.60 45	27.78 43	34.52 36	32.38 40	48.84 42	57.18 45	69.34 47	75.62 50	66.07 48	64.86 48	47.34 43	60.78 47	15.72 45	23.99 45	23.40 44	32.06 45	21.34 44	23.86 44
Nikolai Smolyanskiy, Alexey Kamenev, Stan Birchfield: On the Importance of Stereo for Accurate Depth Estimation: An Efficient Semi-Supervised Deep Neural Network Approach. Arxiv
SGM+DAISY	54.67 46	56.61 46	62.73 47	40.21 46	53.22 46	61.95 47	48.59 48	55.05 46	44.62 32	61.03 44	63.68 45	60.46 45	55.62 45	60.68 47	55.34 45	56.34 46	48.56 48	50.16 46	49.63 48	52.25 46	56.76 48

SPS-STEREO	55.62 47	59.14 47	64.16 48	45.05 47	53.84 48	59.88 46	44.00 47	59.53 47	49.94 37	62.33 46	61.09 43	59.80 44	56.82 46	60.06 46	57.85 46	58.41 47	48.34 47	51.07 47	49.21 47	55.41 47	56.48 47
K. Yamaguchi, D. McAllester, R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014
PWCK	65.90 48	88.63 50	79.40 50	80.32 50	29.96 45	63.85 48	39.45 45	72.07 49	71.26 49	78.74 49	68.99 47	77.82 49	61.77 47	82.81 50	62.64 49	80.17 48	39.88 46	81.90 50	47.37 46	73.54 49	37.51 46

MFMNet_re	72.05 49	79.64 48	64.23 49	52.73 48	84.71 50	77.93 50	68.35 49	63.96 48	66.93 47	61.00 43	71.29 48	63.84 47	68.91 49	72.42 49	61.11 48	82.03 50	74.22 50	80.34 49	74.44 49	83.62 50	89.34 50

DPSimNet_ROB	74.29 50	83.27 49	59.30 46	77.19 49	65.13 49	75.14 49	74.95 50	73.99 50	76.94 50	81.99 50	75.48 49	80.18 50	71.76 50	67.39 48	84.00 50	81.87 49	51.14 49	79.99 48	76.70 50	70.17 48	79.21 49

MEDIAN_ROB	99.19 51	99.84 51	99.62 52	98.49 51	98.51 51	98.58 51	97.81 52	98.80 51	98.56 51	99.36 53	99.49 51	99.56 51	99.06 51	98.35 51	98.31 51	99.99 51	99.63 51	100.00 51	100.00 51	99.81 51	99.95 51

AVERAGE_ROB	99.80 52	99.99 52	99.40 51	100.00 52	100.00 52	98.99 52	97.72 51	100.00 52	100.00 52	100.00 54	100.00 52	100.00 54	100.00 52	100.00 54	100.00 54	100.00 52	100.00 52	100.00 51	100.00 51	100.00 52	100.00 52

DPSMNet_ROB	99.95 53	100.00 53	100.00 53	100.00 52	100.00 52	100.00 53	100.00 53	100.00 52	100.00 52	99.15 52	100.00 52	99.96 52	100.00 52	99.99 52	99.99 52	100.00 52	100.00 52	100.00 51	100.00 51	100.00 52	100.00 52

DGTPSM_ROB	99.95 53	100.00 53	100.00 53	100.00 52	100.00 52	100.00 53	100.00 53	100.00 52	100.00 52	99.14 51	100.00 52	99.96 52	100.00 52	99.99 52	99.99 52	100.00 52	100.00 52	100.00 51	100.00 51	100.00 52	100.00 52

LSM0	100.00 55	100.00 53	100.00 53	100.00 52	100.00 52	100.00 53	100.00 53	100.00 52	100.00 52	100.00 54	100.00 52	100.00 54	100.00 52	100.00 54	100.00 54	100.00 52	100.00 52	100.00 51	100.00 51	100.00 52	100.00 52

DPSM_ROB	100.00 55	100.00 53	100.00 53	100.00 52	100.00 52	100.00 53	100.00 53	100.00 52	100.00 52	100.00 54	100.00 52	100.00 54	100.00 52	100.00 54	100.00 54	100.00 52	100.00 52	100.00 51	100.00 51	100.00 52	100.00 52

DPSM	100.00 55	100.00 53	100.00 53	100.00 52	100.00 52	100.00 53	100.00 53	100.00 52	100.00 52	100.00 54	100.00 52	100.00 54	100.00 52	100.00 54	100.00 54	100.00 52	100.00 52	100.00 51	100.00 51	100.00 52	100.00 52

Low-res two-view benchmark