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 views0.77
1
0.90
28
2.01
10
0.85
25
0.00
1
1.47
3
0.01
2
2.92
1
0.93
1
0.12
1
0.53
3
0.44
1
0.16
4
4.33
9
0.35
1
0.00
1
0.00
1
0.00
1
0.00
1
0.30
25
0.01
2
iResNet_ROBtwo views1.05
2
0.76
23
2.06
11
0.07
6
0.00
1
1.35
2
0.03
4
6.80
26
2.57
16
0.79
2
1.20
7
0.94
2
0.12
3
3.81
4
0.41
2
0.00
1
0.00
1
0.00
1
0.00
1
0.08
13
0.01
2
PSMNet_ROBtwo views1.09
3
0.69
20
2.92
17
0.03
4
0.07
12
1.68
7
0.16
7
3.44
3
1.30
2
0.96
3
0.32
1
2.81
8
0.63
9
4.23
7
2.48
12
0.00
1
0.00
1
0.00
1
0.00
1
0.08
13
0.04
8
ETE_ROBtwo views1.16
4
0.52
14
1.71
5
0.01
1
0.02
8
1.79
8
0.22
10
4.10
11
1.32
3
4.68
14
1.18
6
2.81
8
0.08
2
3.56
1
1.12
5
0.00
1
0.00
1
0.00
1
0.00
1
0.01
2
0.06
11
DLCB_ROBtwo views1.17
5
0.22
4
1.28
1
0.08
7
0.00
1
1.51
4
0.23
11
3.39
2
1.52
7
3.43
9
2.00
12
3.40
14
0.31
7
4.29
8
1.82
9
0.00
1
0.00
1
0.00
1
0.00
1
0.00
1
0.00
1
XPNet_ROBtwo views1.17
5
0.41
9
1.68
4
0.04
5
0.01
6
1.65
6
0.17
8
3.46
5
1.39
4
3.89
11
1.01
5
1.60
3
0.53
8
4.35
10
3.07
16
0.00
1
0.00
1
0.00
1
0.00
1
0.03
5
0.02
4
NOSS_ROBtwo views1.29
7
0.22
4
1.35
3
0.64
23
0.16
18
2.56
10
0.00
1
4.14
13
2.37
12
2.29
6
1.76
11
2.84
10
0.02
1
6.80
17
0.45
3
0.00
1
0.00
1
0.00
1
0.00
1
0.04
9
0.11
16
NCCL2two views1.30
8
0.63
17
2.31
12
0.02
2
0.03
9
1.57
5
3.41
25
3.91
10
1.39
4
3.14
8
0.34
2
3.90
16
0.30
6
3.75
3
1.19
6
0.00
1
0.00
1
0.02
23
0.00
1
0.03
5
0.05
9
LALA_ROBtwo views1.33
9
0.59
16
1.74
6
0.02
2
0.06
11
3.25
16
0.50
13
3.45
4
1.39
4
3.70
10
0.59
4
4.59
17
0.25
5
3.58
2
2.79
14
0.00
1
0.00
1
0.00
1
0.00
1
0.04
9
0.02
4
CBMVpermissivetwo views1.56
10
0.24
6
1.92
7
0.31
16
0.03
9
2.90
13
4.33
28
3.79
6
1.81
9
4.98
15
1.73
9
2.70
6
1.68
15
3.81
4
0.92
4
0.00
1
0.00
1
0.00
1
0.00
1
0.05
12
0.08
14
Konstantinos Batsos, Changjiang Cai, Philippos Mordohai: CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation. Computer Vision and Pattern Recognition (CVPR) 2018
CBMV_ROBtwo views1.70
11
0.11
3
1.98
8
0.46
19
0.00
1
2.85
12
0.91
15
4.27
15
2.21
11
4.02
12
5.00
16
2.97
11
2.32
18
5.40
11
1.35
7
0.00
1
0.00
1
0.00
1
0.00
1
0.03
5
0.06
11
SGM-Foresttwo views1.84
12
0.08
2
1.31
2
0.15
10
0.22
19
3.52
17
2.46
22
3.85
8
2.16
10
5.29
16
3.90
14
3.71
15
1.56
14
6.28
13
1.90
10
0.01
18
0.02
20
0.00
1
0.00
1
0.03
5
0.30
23
Johannes L. Schönberger, Sudipta Sinha, Marc Pollefeys: Learning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching. ECCV 2018
NaN_ROBtwo views2.02
13
0.58
15
2.65
15
0.29
14
0.28
22
3.19
15
5.19
29
3.84
7
2.55
15
5.31
17
1.73
9
2.30
5
1.96
16
6.78
16
3.09
17
0.01
18
0.06
25
0.03
24
0.10
26
0.10
17
0.37
26
PDISCO_ROBtwo views2.21
14
0.78
25
3.43
22
0.96
27
2.04
32
7.04
25
0.09
6
7.36
28
5.72
27
1.21
5
1.67
8
3.16
13
1.28
11
7.52
21
1.37
8
0.00
1
0.00
1
0.00
1
0.00
1
0.34
26
0.13
20
FBW_ROBtwo views2.36
15
0.36
7
3.16
19
0.37
18
0.14
16
4.42
20
0.18
9
6.64
25
3.05
18
2.98
7
2.85
13
4.86
19
1.11
10
11.00
31
4.53
22
0.16
25
0.05
23
0.37
30
0.09
25
0.19
20
0.78
31
CSANtwo views2.70
16
0.75
21
3.41
21
0.18
11
0.12
14
4.18
19
4.05
27
4.11
12
4.36
23
6.03
18
6.24
17
4.60
18
4.74
22
7.29
18
3.43
19
0.02
22
0.03
21
0.01
21
0.02
23
0.18
19
0.22
22
PWC_ROBbinarytwo views2.90
17
1.01
29
6.12
30
0.53
21
0.57
24
1.30
1
0.40
12
3.85
8
4.42
24
7.37
20
15.40
26
3.03
12
1.55
13
6.31
14
5.74
29
0.00
1
0.00
1
0.00
1
0.00
1
0.46
29
0.05
9
PWCDC_ROBbinarytwo views3.06
18
1.38
31
3.81
23
0.14
9
0.00
1
3.14
14
0.02
3
4.40
16
12.35
29
1.08
4
23.69
30
2.01
4
2.01
17
3.87
6
2.43
11
0.24
30
0.00
1
0.00
1
0.00
1
0.61
31
0.08
14
SPS-STEREOcopylefttwo views3.08
19
0.49
13
2.49
13
0.21
12
0.24
20
3.67
18
1.24
18
5.80
22
2.43
14
10.68
23
8.00
18
9.48
21
3.59
21
8.11
23
5.08
26
0.01
18
0.01
17
0.00
1
0.00
1
0.04
9
0.11
16
K. Yamaguchi, D. McAllester, R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014
NVStereoNet_ROBtwo views3.25
20
0.46
11
2.71
16
0.34
17
0.56
23
2.36
9
0.87
14
4.56
17
4.47
25
4.55
13
14.09
25
11.93
22
3.23
20
9.79
27
3.14
18
0.43
31
0.03
21
0.53
31
0.31
32
0.36
28
0.38
27
Nikolai Smolyanskiy, Alexey Kamenev, Stan Birchfield: On the Importance of Stereo for Accurate Depth Estimation: An Efficient Semi-Supervised Deep Neural Network Approach. Arxiv
MDST_ROBtwo views3.50
21
0.07
1
3.94
24
2.66
31
1.46
30
12.94
31
0.97
16
7.32
27
2.42
13
14.82
29
8.08
19
2.74
7
1.51
12
8.44
24
2.48
12
0.00
1
0.00
1
0.00
1
0.00
1
0.01
2
0.11
16
pmcnntwo views3.61
22
0.68
19
5.38
29
0.61
22
1.46
30
2.75
11
1.26
19
5.04
21
1.57
8
8.77
21
11.67
21
18.69
31
2.38
19
6.03
12
5.61
28
0.00
1
0.07
27
0.00
1
0.00
1
0.21
21
0.07
13
SGM_ROBbinarytwo views4.07
23
0.38
8
1.98
8
0.84
24
0.26
21
6.20
24
2.37
21
6.45
23
3.57
20
11.56
24
9.27
20
15.55
28
6.78
23
10.07
28
4.60
23
0.22
26
0.25
30
0.20
27
0.23
30
0.26
22
0.34
25
Heiko Hirschmueller: Stereo processing by semiglobal matching and mutual information. TPAMI 2008, Volume 30(2), pp. 328-341
SGM+DAISYtwo views4.39
24
1.34
30
5.18
27
1.01
28
0.88
26
5.61
23
2.86
24
4.88
19
3.53
19
13.29
26
13.95
24
12.66
24
7.55
24
7.98
22
5.39
27
0.23
29
0.18
28
0.15
26
0.24
31
0.29
23
0.64
28
WCMA_ROBtwo views4.54
25
0.48
12
3.19
20
0.49
20
0.63
25
5.38
22
2.57
23
4.70
18
3.77
21
14.17
28
19.02
27
15.47
27
9.03
27
6.73
15
4.85
24
0.01
18
0.05
23
0.01
21
0.01
22
0.08
13
0.16
21
SANettwo views4.63
26
0.89
27
5.34
28
0.30
15
0.09
13
7.43
27
3.75
26
6.53
24
16.16
32
6.79
19
13.29
23
12.44
23
7.74
25
7.38
19
4.11
21
0.00
1
0.01
17
0.00
1
0.00
1
0.10
17
0.31
24
MSMD_ROBtwo views4.94
27
0.43
10
2.58
14
0.12
8
0.01
6
8.43
29
1.09
17
4.25
14
4.11
22
13.11
25
26.36
32
13.95
26
13.88
30
7.45
20
2.92
15
0.10
24
0.01
17
0.00
1
0.00
1
0.01
2
0.02
4
MeshStereopermissivetwo views5.78
28
0.75
21
3.04
18
0.25
13
0.14
16
7.88
28
2.09
20
9.64
30
5.18
26
20.48
30
19.06
28
22.80
32
9.54
28
9.64
26
4.88
25
0.00
1
0.00
1
0.00
1
0.00
1
0.29
23
0.03
7
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
ELAS_ROBcopylefttwo views7.69
29
0.80
26
4.90
26
1.06
29
1.02
27
10.00
30
9.43
32
9.75
31
14.69
31
21.83
31
20.27
29
17.88
29
16.41
32
13.01
32
10.44
31
0.22
26
0.55
32
0.28
29
0.19
27
0.35
27
0.65
29
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
DispFullNettwo views7.91
30
23.96
34
13.67
32
14.71
33
7.88
33
4.83
21
0.04
5
5.01
20
2.89
17
8.86
22
4.96
15
6.11
20
10.72
29
10.18
29
3.51
20
2.44
32
0.33
31
15.15
33
4.65
33
12.33
33
5.95
33
ELAScopylefttwo views8.05
31
0.76
23
4.02
25
0.95
26
1.02
27
14.85
32
6.99
31
12.80
33
11.73
28
22.43
32
27.75
33
17.88
29
15.19
31
10.96
30
11.31
32
0.22
26
0.56
33
0.23
28
0.22
29
0.50
30
0.71
30
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
PWCKtwo views8.72
32
5.37
32
16.61
33
5.74
32
0.12
14
19.88
33
11.41
33
11.31
32
13.42
30
14.07
27
12.84
22
13.20
25
8.05
26
15.44
33
10.25
30
5.95
33
0.22
29
2.09
32
0.05
24
7.50
32
0.92
32
LE_ROBtwo views12.79
33
0.64
18
10.26
31
2.13
30
1.29
29
7.25
26
5.81
30
9.09
29
49.27
34
57.55
34
24.58
31
23.68
33
31.13
34
9.00
25
23.57
33
0.04
23
0.06
25
0.13
25
0.20
28
0.08
13
0.11
16
DPSimNet_ROBtwo views25.03
34
19.56
33
24.93
34
24.36
34
17.85
34
22.25
34
25.93
34
28.52
34
29.10
33
32.61
33
28.25
34
32.87
34
26.42
33
24.51
34
40.42
34
16.75
34
13.93
34
25.19
34
22.38
34
19.96
34
24.75
34
MEDIAN_ROBtwo views96.83
35
99.41
35
98.66
36
94.75
35
94.23
35
93.08
35
90.54
35
95.61
35
94.75
35
97.65
37
97.73
35
98.30
35
96.57
35
94.51
35
93.74
35
99.78
39
98.24
35
99.99
37
99.89
35
99.48
37
99.67
37
DPSMtwo views99.17
36
100.00
39
100.00
39
96.28
36
98.84
36
100.00
37
100.00
37
100.00
38
100.00
36
100.00
38
100.00
36
100.00
38
100.00
36
100.00
39
100.00
39
93.66
35
100.00
36
100.00
38
100.00
36
96.31
35
98.27
35
DPSM_ROBtwo views99.17
36
100.00
39
100.00
39
96.28
36
98.84
36
100.00
37
100.00
37
100.00
38
100.00
36
100.00
38
100.00
36
100.00
38
100.00
36
100.00
39
100.00
39
93.66
35
100.00
36
100.00
38
100.00
36
96.31
35
98.27
35
AVERAGE_ROBtwo views99.23
38
99.81
36
97.56
35
100.00
40
100.00
38
96.24
36
91.02
36
100.00
38
100.00
36
100.00
38
100.00
36
100.00
38
100.00
36
99.98
38
99.97
36
100.00
40
100.00
36
100.00
38
100.00
36
100.00
40
100.00
39
DGTPSM_ROBtwo views99.45
39
99.94
37
99.98
37
96.58
38
100.00
38
100.00
37
100.00
37
99.81
36
100.00
36
95.86
35
100.00
36
99.32
36
100.00
36
99.79
36
99.99
37
98.34
37
100.00
36
99.86
35
100.00
36
99.52
38
99.99
38
DPSMNet_ROBtwo views99.46
40
99.94
37
99.98
37
96.59
39
100.00
38
100.00
37
100.00
37
99.81
36
100.00
36
95.87
36
100.00
36
99.32
36
100.00
36
99.79
36
99.99
37
98.46
38
100.00
36
99.86
35
100.00
36
99.52
38
100.00
39