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).
Methods with suffix _ROB may participate in the Robust Vision Challenge .
training datasets . The visualizations may not work with mobile browsers.
Coverage : dense
Set : Test
Metric : bad 0.5 [%]
Mask : non-occluded
Method
Info all
DN-CSS_ROB 8.17 1 2.59 7 14.14 13 6.11 5 3.22 10 6.31 1 3.53 2 22.92 2 20.59 2 16.66 2 30.05 4 9.31 1 7.95 1 13.49 1 4.78 1 0.72 17 0.00 1 0.45 12 0.00 1 0.54 7 0.09 1 iResNet 10.26 2 3.10 12 15.72 20 7.35 10 2.13 3 13.86 2 7.07 8 22.80 1 28.01 4 20.10 3 30.40 5 16.79 2 11.06 2 14.46 2 10.29 6 0.39 10 0.05 5 0.00 1 0.07 9 0.73 9 0.73 9 NOSS_ROB 10.99 3 3.81 15 6.88 3 6.69 8 2.30 4 16.06 5 10.94 12 30.06 10 32.41 11 16.10 1 18.72 1 22.78 4 14.21 3 17.52 5 8.62 3 2.35 26 1.84 28 3.32 30 1.57 27 2.06 22 1.52 19 HSM 11.33 4 1.80 2 7.04 5 6.13 6 3.99 12 16.16 6 6.92 7 29.69 8 19.39 1 22.95 5 26.05 2 41.08 20 15.79 6 20.25 11 8.89 4 0.02 1 0.03 3 0.01 4 0.00 1 0.15 1 0.34 4 iResNet_ROB 11.71 5 1.96 5 10.77 8 6.01 4 3.68 11 15.13 4 2.34 1 31.72 12 37.58 19 26.16 7 32.77 10 25.88 7 15.34 4 16.35 3 7.94 2 0.19 4 0.01 2 0.00 1 0.00 1 0.31 2 0.12 2 CBMV_ROB 11.77 6 2.89 9 6.92 4 5.20 1 2.88 7 14.12 3 4.79 4 26.69 3 29.84 6 26.17 8 31.92 8 23.61 5 20.74 17 20.20 10 10.18 5 1.78 25 1.80 27 2.17 27 1.26 25 1.61 17 0.55 7 SGM-Forest 12.92 7 1.87 3 6.61 1 5.68 2 2.05 2 23.19 13 11.30 14 34.24 16 30.77 9 26.54 9 31.99 9 29.98 10 15.74 5 21.17 13 13.36 7 1.05 21 0.33 17 0.84 19 0.01 5 0.71 8 0.97 13 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 8 3.00 10 9.12 7 9.43 16 5.68 20 21.80 10 10.12 10 29.19 7 29.92 7 27.71 11 31.45 7 32.37 12 19.50 14 19.02 8 16.73 11 0.22 5 0.04 4 0.38 11 0.06 8 0.49 5 0.85 10 StereoDRNet-Filtered 13.36 9 2.80 8 10.82 9 7.94 14 3.10 8 18.95 7 4.45 3 28.47 6 29.65 5 29.11 13 41.47 24 24.44 6 17.64 10 23.34 14 21.66 14 0.16 2 0.07 6 0.53 15 0.32 15 0.77 10 1.64 20 CBMV 13.69 10 3.63 13 8.02 6 5.93 3 2.69 5 22.57 11 12.44 15 29.81 9 31.57 10 31.20 20 33.79 12 31.04 11 17.41 9 25.28 17 14.11 8 0.71 16 0.60 24 0.60 16 0.11 10 1.11 14 1.13 14 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 11 3.68 14 19.78 27 6.83 9 4.51 14 21.27 9 14.67 19 27.37 4 30.54 8 27.09 10 40.19 21 38.71 18 18.53 13 17.70 6 19.27 12 0.84 18 0.09 9 0.00 1 0.00 1 0.48 3 0.31 3 StereoDRNet 15.30 12 4.47 18 14.35 16 10.71 21 8.69 31 25.02 17 11.03 13 39.66 25 35.62 15 29.17 14 28.81 3 34.05 14 17.80 11 18.80 7 23.77 20 0.41 11 0.30 16 0.15 7 0.14 11 1.64 18 1.42 17 DRN-Test 15.88 13 3.83 16 14.70 17 9.96 18 8.02 30 28.93 21 14.65 18 42.34 28 38.25 21 28.18 12 34.31 14 29.22 9 18.35 12 20.01 9 22.68 16 0.37 7 0.41 19 0.46 13 0.28 14 1.28 15 1.34 16 DISCO 16.36 14 1.88 4 16.15 22 7.78 13 4.19 13 26.92 18 10.21 11 30.31 11 44.26 26 21.35 4 33.91 13 35.80 17 22.40 18 35.65 31 33.87 33 0.18 3 0.08 7 0.04 5 0.04 6 1.71 19 0.54 6 NaN_ROB 16.51 15 6.51 22 16.22 23 10.53 20 4.64 16 31.76 24 17.78 27 37.00 21 43.37 24 29.68 17 31.29 6 32.67 13 20.32 16 28.00 22 16.06 9 0.38 8 0.41 19 0.32 9 0.41 17 0.87 11 1.94 21 PSMNet_ROB 16.90 16 5.45 20 14.16 14 13.77 27 7.25 28 27.65 19 32.74 33 43.21 31 37.88 20 24.00 6 32.97 11 34.55 15 16.69 7 17.41 4 23.96 22 0.38 8 0.22 12 0.93 21 2.05 29 1.83 20 0.96 12 NCCL2 17.23 17 6.37 21 14.30 15 23.45 34 7.18 27 24.16 15 17.05 25 34.39 17 25.88 3 31.55 21 38.01 16 42.09 22 26.81 24 20.94 12 22.69 17 0.45 12 0.23 13 2.63 28 1.80 28 2.02 21 2.57 26 XPNet_ROB 17.33 18 4.84 19 15.44 19 11.15 22 6.64 24 20.88 8 16.10 23 36.26 20 39.54 22 31.80 22 41.32 23 41.30 21 23.91 19 24.67 15 27.40 28 1.07 22 0.74 25 0.79 18 0.25 12 1.10 13 1.31 15 ETE_ROB 17.67 19 9.48 29 17.66 24 13.60 25 4.57 15 23.03 12 21.09 30 32.49 14 35.24 14 30.49 18 38.06 17 46.18 30 24.49 20 26.40 19 24.21 23 0.46 13 0.24 14 1.14 24 0.84 22 1.52 16 2.13 24 PWCDC_ROB 17.80 20 9.62 30 18.38 25 17.95 30 7.74 29 24.61 16 5.56 6 34.41 18 52.79 32 29.66 16 50.20 30 20.63 3 19.69 15 29.85 24 20.87 13 6.18 30 0.26 15 0.10 6 0.05 7 5.36 30 2.00 22 MDST_ROB 17.92 21 1.60 1 13.68 11 13.88 28 6.44 22 43.05 34 14.86 20 42.74 30 41.66 23 43.25 30 42.85 27 28.43 8 17.28 8 29.35 23 16.06 9 0.88 19 0.11 10 0.63 17 0.47 19 0.50 6 0.64 8 PWC_ROB 18.37 22 10.55 33 25.19 30 10.21 19 6.29 21 23.75 14 4.79 4 35.93 19 48.75 28 32.56 23 44.46 29 34.68 16 24.78 22 30.99 25 25.72 27 1.38 23 0.08 7 1.57 25 0.26 13 2.18 23 3.17 29 LALA_ROB 19.19 23 7.15 26 15.93 21 12.96 24 5.30 19 28.30 20 23.21 31 42.51 29 36.69 16 33.45 25 40.81 22 50.96 32 24.53 21 27.92 21 25.66 26 0.60 14 0.44 21 2.04 26 1.22 24 2.54 24 1.47 18 SGM_ROB 19.52 24 2.12 6 6.79 2 6.59 7 1.33 1 38.20 33 15.74 22 38.13 23 34.07 13 45.71 32 41.91 25 51.41 33 38.10 31 38.77 33 27.80 30 0.61 15 0.36 18 0.51 14 0.33 16 1.03 12 0.90 11 Heiko Hirschmueller: Stereo processing by semiglobal matching and mutual information . TPAMI 2008, Volume 30(2), pp. 328-341 WCMA_ROB 20.02 25 4.30 17 19.72 26 9.47 17 7.00 26 32.71 28 13.99 17 31.97 13 32.48 12 41.51 29 52.00 31 44.09 24 36.14 30 32.43 27 24.29 24 6.19 31 2.79 32 1.09 23 1.10 23 3.95 27 3.13 28 SANet 20.96 26 6.60 23 29.81 32 8.83 15 3.19 9 31.27 23 20.56 29 41.86 27 56.09 36 39.30 27 43.62 28 44.95 26 31.93 28 27.83 20 23.67 19 0.94 20 0.52 23 0.99 22 0.42 18 4.72 29 2.03 23 FBW_ROB 21.00 27 10.23 32 22.72 28 13.71 26 6.77 25 30.49 22 16.46 24 44.58 32 43.57 25 44.25 31 39.59 19 43.27 23 26.12 23 33.96 29 21.80 15 2.58 28 1.84 28 7.14 32 2.78 31 4.00 28 4.06 30 CSAN 21.34 28 9.45 28 23.34 29 20.99 32 4.95 18 32.57 27 34.26 34 38.83 24 49.95 31 36.97 26 39.72 20 44.97 27 31.73 27 26.05 18 23.94 21 1.52 24 0.47 22 0.85 20 1.43 26 2.73 25 2.14 25 MeshStereo 22.27 29 6.87 24 11.15 10 7.69 11 4.87 17 37.70 32 13.06 16 41.64 26 36.95 17 50.92 33 53.41 32 58.12 34 41.93 33 37.73 32 27.62 29 2.52 27 2.37 31 2.99 29 2.07 30 3.09 26 2.60 27 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 30 9.64 31 14.76 18 17.46 29 9.88 32 34.94 30 18.83 28 33.67 15 37.47 18 40.97 28 59.80 34 45.93 28 38.34 32 31.74 26 23.24 18 6.22 32 3.58 34 8.78 33 11.03 35 6.76 31 5.06 32 LE_ROB 22.93 31 3.07 11 14.02 12 7.70 12 2.86 6 31.99 26 17.35 26 28.10 5 67.19 40 70.51 40 55.61 33 49.07 31 47.62 34 24.76 16 36.49 34 0.35 6 0.20 11 0.27 8 0.55 20 0.48 3 0.44 5 testNet 24.40 32 6.97 25 33.14 33 11.16 23 6.54 23 53.33 36 43.32 38 51.28 37 59.37 38 30.89 19 39.36 18 40.35 19 34.30 29 32.57 28 25.09 25 4.65 29 1.62 26 0.37 10 0.66 21 8.51 32 4.44 31 PDISCO_ROB 24.45 33 9.24 27 29.28 31 28.68 35 19.96 35 37.14 31 15.33 21 45.04 33 54.69 34 29.24 15 42.67 26 46.07 29 28.17 25 35.35 30 30.30 32 9.92 33 2.13 30 6.90 31 3.20 32 8.74 33 6.87 33 DispFullNet 30.27 34 31.74 36 58.63 37 30.13 36 28.18 37 31.79 25 8.94 9 37.16 22 54.68 33 32.60 24 37.99 15 44.62 25 28.80 26 41.55 34 27.93 31 11.94 36 3.46 33 35.53 37 12.34 36 30.08 37 17.26 36 ELAS 33.68 35 18.95 34 36.71 35 20.64 31 13.61 33 56.43 37 35.14 35 50.03 36 49.90 29 60.58 34 63.83 38 58.64 36 53.36 35 50.73 36 44.44 36 11.03 34 6.58 35 10.02 34 7.43 34 13.58 35 11.94 35 A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching . ACCV 2010 ELAS_ROB 33.79 36 19.15 35 35.97 34 21.28 33 13.79 34 52.58 35 36.11 36 48.96 35 55.66 35 61.47 37 62.91 36 58.35 35 53.55 36 53.37 37 42.23 35 11.25 35 6.64 36 10.32 35 7.05 33 13.56 34 11.58 34 A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching . ACCV 2010 NVStereoNet_ROB 41.93 37 32.40 37 44.55 36 36.60 37 27.78 36 34.52 29 32.38 32 48.84 34 57.18 37 69.34 39 75.62 42 66.07 40 64.86 40 47.34 35 60.78 39 15.72 37 23.99 37 23.40 36 32.06 37 21.34 36 23.86 37 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 38 56.61 38 62.73 39 40.21 38 53.22 39 61.95 39 48.59 40 55.05 38 44.62 27 61.03 36 63.68 37 60.46 38 55.62 37 60.68 39 55.34 37 56.34 38 48.56 40 50.16 38 49.63 40 52.25 38 56.76 40 SPS-STEREO 55.62 39 59.14 39 64.16 40 45.05 39 53.84 40 59.88 38 44.00 39 59.53 39 49.94 30 62.33 38 61.09 35 59.80 37 56.82 38 60.06 38 57.85 38 58.41 39 48.34 39 51.07 39 49.21 39 55.41 39 56.48 39 K. Yamaguchi, D. McAllester, R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation . ECCV 2014 PWCK 65.90 40 88.63 42 79.40 42 80.32 42 29.96 38 63.85 40 39.45 37 72.07 41 71.26 41 78.74 41 68.99 39 77.82 41 61.77 39 82.81 42 62.64 41 80.17 40 39.88 38 81.90 42 47.37 38 73.54 41 37.51 38 MFMNet_re 72.05 41 79.64 40 64.23 41 52.73 40 84.71 42 77.93 42 68.35 41 63.96 40 66.93 39 61.00 35 71.29 40 63.84 39 68.91 41 72.42 41 61.11 40 82.03 42 74.22 42 80.34 41 74.44 41 83.62 42 89.34 42 DPSimNet_ROB 74.29 42 83.27 41 59.30 38 77.19 41 65.13 41 75.14 41 74.95 42 73.99 42 76.94 42 81.99 42 75.48 41 80.18 42 71.76 42 67.39 40 84.00 42 81.87 41 51.14 41 79.99 40 76.70 42 70.17 40 79.21 41 MEDIAN_ROB 99.19 43 99.84 43 99.62 44 98.49 43 98.51 43 98.58 43 97.81 44 98.80 43 98.56 43 99.36 45 99.49 43 99.56 43 99.06 43 98.35 43 98.31 43 99.99 43 99.63 43 100.00 43 100.00 43 99.81 43 99.95 43 AVERAGE_ROB 99.80 44 99.99 44 99.40 43 100.00 44 100.00 44 98.99 44 97.72 43 100.00 44 100.00 44 100.00 46 100.00 44 100.00 46 100.00 44 100.00 46 100.00 46 100.00 44 100.00 44 100.00 43 100.00 43 100.00 44 100.00 44 DPSMNet_ROB 99.95 45 100.00 45 100.00 45 100.00 44 100.00 44 100.00 45 100.00 45 100.00 44 100.00 44 99.15 44 100.00 44 99.96 44 100.00 44 99.99 44 99.99 44 100.00 44 100.00 44 100.00 43 100.00 43 100.00 44 100.00 44 DGTPSM_ROB 99.95 45 100.00 45 100.00 45 100.00 44 100.00 44 100.00 45 100.00 45 100.00 44 100.00 44 99.14 43 100.00 44 99.96 44 100.00 44 99.99 44 99.99 44 100.00 44 100.00 44 100.00 43 100.00 43 100.00 44 100.00 44 DPSM 100.00 47 100.00 45 100.00 45 100.00 44 100.00 44 100.00 45 100.00 45 100.00 44 100.00 44 100.00 46 100.00 44 100.00 46 100.00 44 100.00 46 100.00 46 100.00 44 100.00 44 100.00 43 100.00 43 100.00 44 100.00 44 DPSM_ROB 100.00 47 100.00 45 100.00 45 100.00 44 100.00 44 100.00 45 100.00 45 100.00 44 100.00 44 100.00 46 100.00 44 100.00 46 100.00 44 100.00 46 100.00 46 100.00 44 100.00 44 100.00 43 100.00 43 100.00 44 100.00 44
