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
sort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysorted by
iResNet_ROBtwo views4.23
4
1.02
9
4.90
9
2.18
15
0.93
12
2.92
1
0.37
3
15.10
17
16.91
21
7.89
4
10.51
5
7.03
3
3.07
3
8.16
4
3.46
3
0.01
2
0.00
1
0.00
1
0.00
1
0.10
8
0.02
1
DN-CSS_ROBtwo views2.69
1
1.40
16
5.34
10
2.31
18
0.75
7
3.14
2
0.06
1
6.11
1
3.87
1
5.34
2
12.18
6
2.34
1
1.22
1
7.84
1
1.48
1
0.03
10
0.00
1
0.00
1
0.00
1
0.35
20
0.03
2
MSMD_ROBtwo views9.28
21
1.09
11
4.65
8
1.58
6
0.39
2
16.52
23
4.41
10
13.60
12
14.87
16
22.34
22
39.89
29
25.67
21
20.71
25
12.42
16
6.98
11
0.34
21
0.03
10
0.00
1
0.00
1
0.05
4
0.09
3
CBMV_ROBtwo views4.14
3
0.52
4
3.14
3
1.30
3
0.77
8
6.92
7
1.97
7
10.11
3
9.58
6
8.92
5
14.20
11
7.12
5
5.90
12
8.65
6
3.50
4
0.01
2
0.05
13
0.00
1
0.00
1
0.04
3
0.09
3
DLCB_ROBtwo views4.51
5
0.91
7
3.78
6
2.19
16
1.07
15
6.28
4
3.09
8
9.78
2
7.72
3
10.65
8
12.97
7
13.91
11
3.71
4
8.72
7
5.30
7
0.00
1
0.00
1
0.00
1
0.00
1
0.03
2
0.10
5
PSMNet_ROBtwo views5.02
7
1.63
20
6.03
13
1.90
12
1.83
20
9.57
11
6.35
17
15.58
18
7.23
2
6.15
3
10.48
4
12.22
10
4.16
5
8.02
3
8.71
17
0.02
5
0.01
6
0.01
15
0.10
22
0.20
15
0.12
6
MDST_ROBtwo views8.37
18
0.32
1
9.03
21
4.18
24
2.42
22
26.86
30
6.14
13
19.36
25
13.52
12
27.09
25
22.75
18
9.47
6
4.74
7
15.06
22
6.34
8
0.02
5
0.02
8
0.00
1
0.00
1
0.02
1
0.13
7
NOSS_ROBtwo views3.30
2
0.46
3
2.62
1
2.08
14
1.01
14
5.60
3
0.74
4
10.37
4
11.48
9
5.15
1
8.43
1
5.67
2
1.73
2
7.97
2
2.34
2
0.02
5
0.06
15
0.00
1
0.00
1
0.07
6
0.14
8
pmcnntwo views7.72
15
1.27
14
9.42
23
2.91
22
3.14
23
9.44
9
6.23
14
12.56
10
16.51
20
14.53
16
24.08
19
27.44
24
8.49
16
9.32
10
8.44
15
0.06
15
0.08
20
0.00
1
0.00
1
0.30
17
0.15
9
LE_ROBtwo views16.73
31
1.28
15
11.61
27
3.72
23
1.65
18
16.67
24
9.17
24
14.39
13
55.91
33
63.81
33
40.86
31
35.94
29
37.73
33
14.24
21
26.87
32
0.05
13
0.10
21
0.13
23
0.22
24
0.12
11
0.15
9
PWCDC_ROBbinarytwo views7.92
16
3.17
28
7.48
19
5.73
27
4.40
24
10.45
16
0.35
2
14.52
15
28.19
27
10.36
7
31.27
24
7.04
4
9.14
17
13.22
19
8.78
18
2.74
27
0.02
8
0.00
1
0.00
1
1.31
25
0.17
11
CBMVpermissivetwo views5.35
8
0.91
7
3.67
5
1.62
7
0.44
3
10.09
14
7.19
22
12.49
8
12.33
11
12.22
11
14.69
12
10.93
7
6.48
13
8.51
5
4.96
6
0.02
5
0.15
23
0.00
1
0.00
1
0.17
14
0.17
11
Konstantinos Batsos, Changjiang Cai, Philippos Mordohai: CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation. Computer Vision and Pattern Recognition (CVPR) 2018
MeshStereopermissivetwo views11.52
25
1.52
17
4.55
7
1.89
11
1.46
17
19.87
26
5.11
11
20.66
27
15.91
19
32.67
29
34.51
27
39.34
32
21.15
26
18.74
25
12.10
24
0.11
18
0.06
15
0.01
15
0.00
1
0.45
21
0.22
13
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
NCCL2two views5.88
10
1.59
18
5.44
11
1.87
10
0.92
11
9.55
10
11.55
30
12.11
7
9.94
7
9.67
6
8.85
2
22.28
19
7.41
15
8.78
8
7.17
12
0.01
2
0.00
1
0.03
18
0.00
1
0.13
12
0.23
14
WCMA_ROBtwo views9.21
20
0.87
6
7.37
18
2.54
20
2.13
21
13.59
20
5.80
12
11.64
6
14.01
13
24.43
23
32.99
25
27.09
23
18.02
21
12.51
17
9.85
21
0.81
24
0.07
18
0.01
15
0.01
15
0.16
13
0.23
14
XPNet_ROBtwo views6.03
12
1.22
12
5.61
12
2.56
21
0.90
10
6.32
5
7.07
20
12.92
11
8.30
5
14.76
17
15.13
15
19.84
15
6.66
14
10.36
12
8.58
16
0.02
5
0.04
11
0.00
1
0.03
19
0.11
10
0.24
16
LALA_ROBtwo views6.58
13
1.80
22
6.25
14
1.26
2
0.94
13
10.08
13
9.02
23
16.00
19
11.51
10
12.74
13
13.02
8
24.77
20
5.25
9
10.56
13
8.02
13
0.04
11
0.05
13
0.00
1
0.02
16
0.10
8
0.25
17
PWC_ROBbinarytwo views8.24
17
3.13
27
12.74
28
2.43
19
4.43
25
7.51
8
1.22
5
16.63
21
19.24
23
16.08
18
28.29
22
13.99
12
10.16
19
13.63
20
14.06
26
0.42
22
0.00
1
0.05
21
0.00
1
0.59
22
0.27
18
ETE_ROBtwo views5.80
9
1.77
21
6.33
16
1.44
5
0.78
9
6.43
6
6.90
19
12.53
9
8.08
4
12.93
14
14.89
13
21.13
18
5.87
11
9.83
11
6.57
10
0.04
11
0.01
6
0.00
1
0.02
16
0.08
7
0.33
19
SGM_ROBbinarytwo views10.08
23
0.60
5
3.42
4
2.30
17
0.32
1
19.41
25
6.33
16
18.95
22
14.64
14
25.14
24
24.32
20
33.34
27
18.79
23
19.86
26
12.55
25
0.25
20
0.26
25
0.22
25
0.24
25
0.34
19
0.40
20
Heiko Hirschmueller: Stereo processing by semiglobal matching and mutual information. TPAMI 2008, Volume 30(2), pp. 328-341
SGM-Foresttwo views4.96
6
0.32
1
2.84
2
1.21
1
0.64
4
10.23
15
6.64
18
11.55
5
10.98
8
10.94
9
13.59
9
11.65
8
4.30
6
8.94
9
4.63
5
0.11
18
0.04
11
0.00
1
0.00
1
0.05
4
0.46
21
Johannes L. Schönberger, Sudipta Sinha, Marc Pollefeys: Learning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching. ECCV 2018
CSANtwo views7.62
14
1.60
19
6.56
17
1.83
9
0.66
5
12.40
17
10.52
29
14.45
14
21.32
24
14.19
15
15.98
16
17.84
13
13.02
20
12.32
15
8.38
14
0.09
17
0.07
18
0.03
18
0.04
20
0.33
18
0.67
22
NaN_ROBtwo views6.00
11
1.24
13
6.29
15
1.34
4
1.68
19
9.60
12
10.31
27
15.09
16
15.79
18
12.62
12
8.95
3
11.67
9
5.83
10
11.78
14
6.41
9
0.05
13
0.13
22
0.08
22
0.20
23
0.22
16
0.79
23
SANettwo views10.64
24
1.86
23
10.91
25
1.76
8
0.71
6
14.62
22
9.23
25
19.18
23
37.14
31
19.22
20
27.96
21
25.86
22
19.11
24
13.02
18
10.63
22
0.08
16
0.06
15
0.03
18
0.02
16
0.62
23
0.81
24
FBW_ROBtwo views8.50
19
1.03
10
7.98
20
1.93
13
1.28
16
13.10
19
6.23
14
22.50
28
18.98
22
18.82
19
14.91
14
19.06
14
10.04
18
18.41
24
9.83
20
0.62
23
0.22
24
1.82
28
0.82
27
0.99
24
1.36
25
ELAS_ROBcopylefttwo views16.54
29
2.26
26
10.09
24
5.50
26
4.46
26
28.28
31
16.72
32
25.55
31
33.54
30
40.19
30
40.30
30
36.68
30
30.03
31
29.40
32
20.61
29
0.98
26
1.21
27
0.86
27
0.70
26
1.39
26
2.16
26
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
ELAScopylefttwo views16.72
30
2.14
25
9.23
22
4.92
25
4.53
27
32.66
32
15.11
31
27.40
32
28.68
29
40.27
31
44.90
32
38.33
31
30.50
32
26.44
31
21.94
31
0.88
25
1.23
28
0.67
26
0.89
28
1.49
27
2.18
27
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
PDISCO_ROBtwo views9.62
22
1.99
24
11.51
26
9.88
30
9.61
30
21.48
27
3.83
9
19.33
24
28.49
28
11.27
10
14.17
10
19.92
16
5.02
8
16.35
23
9.18
19
5.28
29
0.41
26
0.14
24
0.09
21
2.05
28
2.36
28
SPS-STEREOcopylefttwo views15.04
26
6.23
29
13.21
30
11.34
31
11.65
32
23.30
28
7.15
21
24.16
30
15.65
17
31.78
28
29.19
23
31.62
25
21.32
27
24.62
29
19.50
28
7.59
31
4.19
32
3.22
29
1.48
29
6.99
31
6.54
29
K. Yamaguchi, D. McAllester, R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014
SGM+DAISYtwo views15.62
27
7.26
31
19.28
31
8.94
29
10.11
31
26.25
29
10.49
28
19.36
25
14.65
15
30.64
27
33.59
26
33.00
26
22.32
28
24.96
30
16.42
27
7.90
32
6.25
33
4.51
30
3.37
30
5.86
30
7.20
30
PWCKtwo views30.53
33
44.32
33
47.25
34
29.76
33
7.23
28
40.78
33
27.10
33
44.73
33
44.32
32
47.31
32
36.37
28
47.16
33
26.05
30
41.26
33
31.87
33
21.83
33
4.03
31
29.50
32
4.67
31
27.17
33
7.80
31
NVStereoNet_ROBtwo views16.04
28
6.75
30
12.90
29
6.37
28
7.42
29
12.89
18
9.74
26
22.78
29
25.12
25
30.32
26
46.19
33
34.37
28
25.38
29
21.48
27
21.38
30
5.94
30
3.10
30
6.07
31
10.09
33
4.01
29
8.54
32
Nikolai Smolyanskiy, Alexey Kamenev, Stan Birchfield: On the Importance of Stereo for Accurate Depth Estimation: An Efficient Semi-Supervised Deep Neural Network Approach. Arxiv
DispFullNettwo views17.47
32
26.01
32
33.98
32
22.58
32
20.86
33
13.84
21
1.28
6
16.50
20
26.27
26
19.97
21
17.17
17
20.52
17
18.49
22
22.86
28
10.76
23
5.13
28
2.83
29
30.72
33
7.72
32
20.86
32
11.01
33
DPSimNet_ROBtwo views53.45
34
64.73
34
44.39
33
53.97
34
45.39
34
53.66
34
54.83
34
55.15
34
57.87
34
64.16
34
50.83
34
63.40
34
53.34
34
46.45
34
65.81
34
63.13
34
26.54
34
57.94
34
51.11
34
45.52
34
50.69
34
MEDIAN_ROBtwo views98.41
35
99.70
35
99.30
36
97.09
35
97.02
35
96.89
35
95.77
36
97.66
35
97.28
35
98.79
37
98.94
35
99.18
35
98.14
35
96.89
35
96.88
35
99.96
37
99.16
35
100.00
35
99.99
35
99.69
35
99.88
35
DPSMtwo views99.95
39
100.00
37
100.00
39
99.76
36
100.00
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
38
100.00
39
99.21
35
100.00
36
100.00
35
100.00
36
99.99
36
99.95
36
DPSM_ROBtwo views99.95
39
100.00
37
100.00
39
99.76
36
100.00
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
38
100.00
39
99.21
35
100.00
36
100.00
35
100.00
36
99.99
36
99.95
36
AVERAGE_ROBtwo views99.62
36
99.95
36
98.81
35
100.00
40
100.00
36
98.08
36
95.47
35
100.00
38
100.00
36
100.00
38
100.00
36
100.00
38
100.00
36
100.00
38
99.99
36
100.00
39
100.00
36
100.00
35
100.00
36
100.00
38
100.00
38
DGTPSM_ROBtwo views99.90
37
100.00
37
99.99
37
99.99
38
100.00
36
100.00
37
100.00
37
99.97
36
100.00
36
98.35
35
100.00
36
99.84
36
100.00
36
99.98
36
99.99
36
99.99
38
100.00
36
100.00
35
100.00
36
100.00
38
100.00
38
DPSMNet_ROBtwo views99.91
38
100.00
37
99.99
37
99.99
38
100.00
36
100.00
37
100.00
37
99.98
37
100.00
36
98.35
35
100.00
36
99.84
36
100.00
36
99.98
36
99.99
36
100.00
39
100.00
36
100.00
35
100.00
36
100.00
38
100.00
38