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):

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 Infoalllakes. 1llakes. 1ssand box 1lsand box 1sstora. room 1lstora. room 1sstora. room 2lstora. room 2sstora. room 2 1lstora. room 2 1sstora. room 2 2lstora. room 2 2sstora. room 3lstora. room 3stunnel 1ltunnel 1stunnel 2ltunnel 2stunnel 3ltunnel 3s
sorted bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
DN-CSS_ROBtwo views0.22
1
0.25
14
0.47
7
0.24
6
0.14
3
0.25
1
0.12
1
0.40
1
0.33
1
0.29
1
0.42
2
0.22
1
0.20
1
0.33
1
0.19
1
0.07
1
0.06
1
0.11
3
0.11
5
0.11
4
0.07
1
iResNet_ROBtwo views0.25
2
0.19
5
0.40
3
0.20
1
0.12
1
0.30
2
0.16
2
0.55
9
0.53
13
0.38
4
0.43
3
0.37
3
0.26
2
0.38
2
0.22
2
0.08
2
0.06
1
0.06
1
0.04
1
0.09
1
0.09
3
DLCB_ROBtwo views0.28
3
0.16
2
0.34
1
0.27
13
0.16
7
0.38
6
0.25
8
0.48
2
0.43
3
0.46
7
0.46
5
0.51
10
0.33
5
0.53
5
0.33
6
0.10
4
0.10
6
0.11
3
0.11
5
0.10
2
0.09
3
NOSS_ROBtwo views0.31
4
0.20
7
0.35
2
0.24
6
0.16
7
0.32
3
0.19
3
0.52
5
0.48
7
0.33
2
0.36
1
0.42
4
0.28
3
0.93
20
0.24
3
0.19
23
0.20
26
0.24
25
0.22
25
0.17
16
0.17
20
CBMVpermissivetwo views0.33
5
0.21
11
0.54
12
0.23
4
0.13
2
0.42
11
0.33
15
0.53
6
0.48
7
0.52
12
0.49
8
0.50
9
0.41
15
0.56
8
0.31
4
0.15
13
0.16
18
0.18
19
0.16
18
0.13
8
0.13
8
Konstantinos Batsos, Changjiang Cai, Philippos Mordohai: CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation. Computer Vision and Pattern Recognition (CVPR) 2018
PSMNet_ROBtwo views0.33
5
0.24
13
0.54
12
0.31
20
0.21
17
0.42
11
0.43
23
0.59
17
0.47
5
0.37
3
0.44
4
0.49
6
0.31
4
0.64
10
0.43
11
0.14
11
0.10
6
0.15
12
0.14
13
0.13
8
0.11
6
CBMV_ROBtwo views0.33
5
0.18
4
0.53
11
0.21
2
0.14
3
0.33
4
0.20
6
0.51
3
0.45
4
0.51
10
0.55
13
0.45
5
0.42
19
0.71
12
0.32
5
0.18
20
0.19
25
0.23
24
0.21
23
0.14
10
0.15
14
XPNet_ROBtwo views0.33
5
0.20
7
0.43
4
0.27
13
0.18
11
0.37
5
0.31
10
0.55
9
0.50
10
0.51
10
0.53
11
0.58
13
0.37
10
0.63
9
0.45
13
0.17
16
0.12
12
0.13
9
0.12
9
0.15
11
0.14
10
ETE_ROBtwo views0.34
9
0.26
16
0.45
5
0.29
15
0.18
11
0.40
9
0.37
17
0.57
13
0.47
5
0.50
9
0.50
9
0.62
15
0.36
8
0.55
6
0.38
8
0.13
7
0.10
6
0.14
10
0.12
9
0.16
13
0.16
17
NCCL2two views0.35
10
0.26
16
0.49
9
0.36
23
0.22
21
0.41
10
0.41
21
0.53
6
0.42
2
0.47
8
0.46
5
0.61
14
0.39
12
0.55
6
0.37
7
0.16
14
0.13
16
0.21
21
0.21
23
0.16
13
0.16
17
LALA_ROBtwo views0.36
11
0.25
14
0.46
6
0.30
16
0.21
17
0.47
15
0.39
19
0.61
19
0.51
12
0.52
12
0.51
10
0.69
19
0.36
8
0.50
3
0.43
11
0.17
16
0.11
9
0.16
14
0.14
13
0.17
16
0.15
14
SGM-Foresttwo views0.36
11
0.17
3
0.47
7
0.23
4
0.16
7
0.45
14
0.41
21
0.55
9
0.48
7
0.52
12
0.60
15
0.52
12
0.41
15
0.85
17
0.50
17
0.17
16
0.17
20
0.17
17
0.15
16
0.15
11
0.15
14
Johannes L. Schönberger, Sudipta Sinha, Marc Pollefeys: Learning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching. ECCV 2018
PWCDC_ROBbinarytwo views0.38
13
0.30
23
0.60
15
0.33
22
0.20
14
0.42
11
0.19
3
0.58
14
0.89
28
0.42
5
1.26
25
0.36
2
0.34
6
0.50
3
0.38
8
0.18
20
0.11
9
0.11
3
0.09
3
0.19
23
0.13
8
PWC_ROBbinarytwo views0.38
13
0.29
22
0.69
21
0.25
9
0.20
14
0.38
6
0.19
3
0.58
14
0.67
20
0.57
17
0.85
20
0.51
10
0.40
13
0.71
12
0.52
19
0.13
7
0.09
5
0.14
10
0.10
4
0.17
16
0.14
10
NaN_ROBtwo views0.41
15
0.28
20
0.62
17
0.30
16
0.19
13
0.51
16
0.47
26
0.58
14
0.59
16
0.56
16
0.47
7
0.49
6
0.41
15
1.21
27
0.64
24
0.12
6
0.18
24
0.12
6
0.13
11
0.11
4
0.14
10
FBW_ROBtwo views0.43
16
0.26
16
0.54
12
0.31
20
0.20
14
0.51
16
0.32
11
0.70
23
0.60
17
0.59
18
0.55
13
0.65
17
0.41
15
1.40
32
0.51
18
0.13
7
0.17
20
0.21
21
0.16
18
0.17
16
0.18
22
PDISCO_ROBtwo views0.43
16
0.30
23
0.67
19
0.43
27
0.36
27
0.67
23
0.32
11
0.72
25
0.76
25
0.43
6
0.53
11
0.63
16
0.40
13
0.66
11
0.47
14
0.21
24
0.12
12
0.21
21
0.19
21
0.25
25
0.20
25
MDST_ROBtwo views0.48
18
0.14
1
0.95
26
0.30
16
0.21
17
1.33
32
0.32
11
0.77
27
0.56
15
1.06
26
0.71
17
0.49
6
0.35
7
1.26
29
0.38
8
0.13
7
0.11
9
0.16
14
0.13
11
0.12
7
0.12
7
pmcnntwo views0.50
19
0.20
7
0.78
22
0.24
6
0.26
24
0.39
8
0.30
9
0.51
3
0.50
10
0.54
15
1.23
24
2.52
33
0.37
10
0.77
15
0.95
28
0.08
2
0.06
1
0.06
1
0.05
2
0.10
2
0.08
2
CSANtwo views0.50
19
0.35
26
0.78
22
0.36
23
0.23
22
0.56
20
0.59
28
0.61
19
0.70
22
0.64
19
0.78
18
0.65
17
0.60
20
1.38
31
0.62
22
0.21
24
0.17
20
0.20
20
0.20
22
0.18
22
0.18
22
SGM_ROBbinarytwo views0.50
19
0.19
5
0.50
10
0.25
9
0.15
5
0.69
24
0.39
19
0.68
22
0.82
27
0.95
23
0.84
19
1.13
24
0.76
23
1.16
26
0.60
21
0.16
14
0.16
18
0.16
14
0.16
18
0.16
13
0.17
20
Heiko Hirschmueller: Stereo processing by semiglobal matching and mutual information. TPAMI 2008, Volume 30(2), pp. 328-341
WCMA_ROBtwo views0.51
22
0.21
11
0.65
18
0.25
9
0.21
17
0.58
21
0.32
11
0.54
8
0.55
14
0.95
23
1.40
28
1.28
28
0.81
25
0.73
14
0.62
22
0.18
20
0.15
17
0.15
12
0.15
16
0.19
23
0.19
24
SANettwo views0.53
23
0.28
20
0.96
27
0.26
12
0.15
5
0.69
24
0.44
24
0.67
21
1.34
32
0.67
20
0.98
22
0.94
21
0.71
22
0.89
19
0.76
25
0.14
11
0.12
12
0.12
6
0.11
5
0.17
16
0.16
17
MeshStereopermissivetwo views0.58
24
0.27
19
0.67
19
0.22
3
0.17
10
0.66
22
0.37
17
0.78
28
0.61
18
1.47
32
1.30
26
1.65
30
0.79
24
1.12
25
0.59
20
0.17
16
0.17
20
0.17
17
0.14
13
0.17
16
0.14
10
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_ROBtwo views0.60
25
0.33
25
0.61
16
0.30
16
0.25
23
0.86
28
0.35
16
0.55
9
0.67
20
1.10
27
1.49
29
1.76
31
0.97
29
0.88
18
0.49
16
0.23
26
0.21
27
0.27
26
0.27
26
0.25
25
0.24
26
DispFullNettwo views0.66
26
0.89
32
1.59
31
0.77
32
1.21
33
0.51
16
0.23
7
0.59
17
0.72
23
0.69
21
0.61
16
0.69
19
0.91
28
0.79
16
0.48
15
0.27
28
0.12
12
0.73
32
0.30
29
0.65
32
0.40
29
NVStereoNet_ROBtwo views0.67
27
0.49
29
0.83
24
0.48
28
0.40
28
0.51
16
0.46
25
0.70
23
0.77
26
0.84
22
1.72
32
1.02
23
0.83
26
1.23
28
0.79
26
0.32
30
0.38
30
0.40
29
0.46
30
0.36
29
0.41
30
Nikolai Smolyanskiy, Alexey Kamenev, Stan Birchfield: On the Importance of Stereo for Accurate Depth Estimation: An Efficient Semi-Supervised Deep Neural Network Approach. Arxiv
SPS-STEREOcopylefttwo views0.69
28
0.61
30
0.98
28
0.52
31
0.57
31
0.74
26
0.50
27
0.78
28
0.62
19
0.95
23
0.86
21
0.94
21
0.70
21
1.01
22
0.87
27
0.58
32
0.51
32
0.50
30
0.50
31
0.55
31
0.58
32
K. Yamaguchi, D. McAllester, R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014
ELAS_ROBcopylefttwo views0.74
29
0.36
27
1.00
29
0.37
26
0.33
25
0.88
29
0.93
32
0.83
30
1.08
30
1.35
29
1.33
27
1.24
26
1.33
32
1.06
23
0.95
28
0.27
28
0.25
28
0.29
28
0.27
26
0.30
27
0.30
27
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
ELAScopylefttwo views0.74
29
0.36
27
0.85
25
0.36
23
0.33
25
1.36
33
0.77
29
0.93
32
0.92
29
1.41
31
1.53
31
1.16
25
1.17
30
0.95
21
1.03
30
0.26
27
0.25
28
0.28
27
0.28
28
0.31
28
0.30
27
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
SGM+DAISYtwo views0.87
31
0.66
31
1.30
30
0.51
30
0.60
32
1.03
30
0.84
30
0.76
26
0.73
24
1.39
30
1.51
30
1.31
29
1.22
31
1.11
24
1.08
32
0.57
31
0.53
33
0.51
31
0.51
32
0.54
30
0.61
33
PWCKtwo views1.00
32
1.17
33
1.70
32
0.91
33
0.41
29
1.19
31
0.92
31
1.10
33
1.14
31
1.16
28
1.14
23
1.25
27
0.88
27
1.75
33
1.04
31
0.87
33
0.50
31
0.87
33
0.53
33
0.96
33
0.52
31
LE_ROBtwo views1.76
33
0.20
7
2.68
33
0.48
28
0.52
30
0.78
27
0.96
33
0.84
31
6.61
34
7.40
36
2.08
33
2.08
32
4.83
33
1.27
30
3.79
33
0.10
4
0.08
4
0.12
6
0.11
5
0.11
4
0.10
5
DPSimNet_ROBtwo views4.34
34
4.23
34
6.89
34
3.67
34
3.68
34
4.75
34
5.21
34
2.67
34
3.68
33
5.82
35
3.95
34
5.57
34
6.72
34
3.46
34
4.48
34
4.05
36
2.88
34
4.68
36
3.12
34
3.69
34
3.62
34
DGTPSM_ROBtwo views8.34
35
5.10
35
10.37
37
5.31
35
10.18
35
8.33
35
23.60
37
6.06
35
13.41
35
4.90
33
10.87
35
5.65
35
10.44
35
6.17
35
12.59
35
3.74
34
7.55
35
3.69
34
7.26
37
4.14
35
7.46
35
DPSMNet_ROBtwo views8.40
36
5.11
36
10.49
38
5.58
36
10.25
36
8.34
36
23.62
38
6.07
36
13.45
36
4.93
34
10.88
36
5.66
36
10.44
35
6.24
36
12.64
36
3.98
35
7.61
36
3.76
35
7.30
38
4.20
36
7.51
36
DPSMtwo views11.49
37
9.87
37
10.35
35
11.13
37
11.31
37
19.11
39
27.51
39
13.37
37
14.21
37
10.31
37
11.06
37
10.96
37
11.27
37
11.96
37
13.59
37
6.78
37
8.19
37
6.03
37
7.09
35
7.93
37
7.73
37
DPSM_ROBtwo views11.49
37
9.87
37
10.35
35
11.13
37
11.31
37
19.11
39
27.51
39
13.37
37
14.21
37
10.31
37
11.06
37
10.96
37
11.27
37
11.96
37
13.59
37
6.78
37
8.19
37
6.03
37
7.09
35
7.93
37
7.73
37
MEDIAN_ROBtwo views21.21
39
24.62
39
23.47
39
19.58
39
19.65
39
13.22
37
10.96
35
17.88
39
17.00
39
22.14
39
22.02
39
20.86
39
20.36
39
21.06
39
19.71
39
25.63
39
24.13
39
26.21
39
25.20
39
25.17
39
25.38
39
AVERAGE_ROBtwo views25.43
40
29.06
40
27.24
40
24.63
40
24.20
40
17.73
38
12.61
36
22.29
40
21.39
40
26.79
40
26.16
40
25.20
40
24.64
40
25.07
40
23.53
40
29.96
40
28.40
40
30.60
40
29.58
40
29.72
40
29.84
40