This table lists the benchmark results for the low-res many-view scenario. The following metrics are evaluated:

(*) For exact definitions, detailing how potentially incomplete ground truth is taken into account, see our paper.

The datasets are grouped into different categories, and result averages are computed for a category and method if results of the method are available for all datasets within the category. Note that the category "all" includes both the high-res multi-view and the low-res many-view scenarios.

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

Click a dataset result cell to show a visualization of the reconstruction. For training datasets, ground truth and accuracy / completeness visualizations are also available. The visualizations may not work with mobile browsers.




Method Infoalllow-res
many-view
indooroutdoorlakesidesand boxstorage roomstorage room 2tunnel
sort bysorted bysort bysort bysort bysort bysort bysort bysort by
DeepC-MVS_fast98.69 196.77 195.30 197.74 198.24 297.85 392.74 297.86 197.13 1
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepPCF-MVS98.18 396.68 295.10 297.74 198.30 197.78 592.41 397.79 297.12 2
DeepC-MVS98.35 296.09 394.20 1097.34 397.94 497.39 1090.85 1397.56 396.71 4
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
COLMAP(SR)96.07 494.39 797.19 596.86 897.92 291.54 1097.24 796.79 3
TAPA-MVS(SR)95.82 594.80 396.49 996.19 1297.58 992.04 497.56 395.71 12
ACMM97.58 595.68 694.49 696.48 1096.12 1397.20 1392.04 496.94 1096.10 9
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PCF-MVS97.08 1395.68 693.36 1497.23 498.16 396.93 1491.43 1195.29 1996.61 5
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMH+97.24 995.61 893.77 1396.83 696.69 1097.71 690.70 1496.85 1196.09 10
TAPA-MVS97.07 1495.48 994.55 496.10 1295.68 1697.30 1191.73 897.37 595.32 15
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
COLMAP(base)95.42 1094.03 1296.35 1195.96 1496.92 1591.10 1296.97 896.16 8
PLCcopyleft97.94 495.35 1194.28 996.06 1395.26 1996.59 1691.61 996.96 996.34 7
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMH97.28 795.29 1293.25 1596.65 796.32 1197.93 189.82 1696.68 1495.69 13
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
ACMP97.20 1095.11 1393.01 1896.52 895.68 1697.26 1289.74 1796.27 1696.61 5
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
LTVRE_ROB97.16 1195.10 1494.13 1195.75 1495.69 1595.73 2092.87 195.39 1895.83 11
Andreas Kuhn, Heiko Hirschmüller, Daniel Scharstein, Helmut Mayer: A TV Prior for High-Quality Scalable Multi-View Stereo Reconstruction. International Journal of Computer Vision 2016
GSE95.00 1594.33 895.45 1794.81 2096.53 1891.98 696.68 1495.02 17
A-TVSNet + Gipumacopyleft94.96 1694.54 595.24 1995.31 1896.58 1791.82 797.27 693.83 18
BP-MVSNet94.57 1793.20 1695.49 1596.77 994.34 2889.67 1896.74 1295.35 14
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
COLMAP_ROBcopyleft97.56 694.43 1893.09 1795.33 1894.55 2196.37 1989.45 1996.72 1395.07 16
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
3Dnovator97.25 892.47 1987.97 2795.47 1697.93 597.85 386.02 2389.92 2790.64 22
3Dnovator+97.12 1292.17 2087.57 2895.24 1997.49 697.68 785.24 2689.91 2890.54 23
LPCS92.07 2192.93 1991.49 2690.32 3093.29 3090.66 1595.20 2090.87 21
OpenMVScopyleft96.50 1591.77 2286.65 3095.19 2197.38 797.66 884.60 2988.69 3290.53 24
R-MVSNet90.93 2389.80 2191.68 2590.81 2994.15 2987.67 2091.93 2590.07 25
PVSNet_094.43 1790.46 2485.77 3193.59 2293.82 2395.33 2384.62 2886.92 3491.61 20
test_112689.66 2588.06 2690.73 2893.40 2495.29 2484.99 2791.13 2683.50 33
test_120589.53 2689.08 2489.83 3192.63 2695.71 2185.34 2592.82 2481.15 39
CPR_FA89.35 2788.37 2590.00 3089.38 3287.69 3887.29 2289.46 3092.93 19
PVSNet96.02 1689.22 2884.53 3592.35 2393.29 2594.60 2785.85 2483.20 3789.17 26
CIDER88.55 2983.13 3792.16 2493.95 2295.61 2283.79 3282.47 3886.92 28
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
ANet-0.7587.89 3089.58 2286.76 3284.37 4490.29 3484.51 3094.65 2185.61 31
OpenMVS_ROBcopyleft92.34 1887.46 3181.75 3891.27 2792.55 2795.21 2581.21 3682.29 3986.05 29
P-MVSNet85.35 3291.57 2081.21 4178.76 5182.87 4687.56 2195.58 1782.00 36
ANet85.16 3385.20 3385.14 3584.37 4490.29 3482.00 3488.39 3380.78 41
unMVSv183.89 3484.37 3683.58 3784.00 4885.63 4083.17 3385.56 3581.11 40
metmvs_fine83.50 3584.90 3482.56 3984.72 3981.80 4981.09 3788.72 3181.17 38
A1Net82.90 3689.39 2378.57 5085.19 3663.18 6384.36 3194.42 2287.34 27
RMVSNet80.62 3786.71 2976.57 5284.14 4678.09 5180.22 3893.20 2367.48 58
Pnet_fast80.38 3865.52 5590.29 2990.03 3194.91 2660.54 5470.51 5785.93 30
AttMVS80.32 3978.27 3981.68 4078.98 5086.16 3979.76 3976.79 5279.89 42
Pnet-new-79.46 4069.54 5186.08 3384.76 3890.31 3365.13 4873.95 5483.16 34
CasMVSNet(SR_A)79.00 4169.56 5085.29 3485.95 3492.82 3162.08 5177.05 4877.11 43
hgnet78.82 4276.20 4180.57 4584.50 4082.32 4775.56 4176.84 4974.88 47
DPSNet78.82 4276.20 4180.57 4584.50 4082.32 4775.56 4176.84 4974.88 47
Pnet-blend++78.13 4473.62 4581.13 4284.46 4288.20 3666.50 4580.74 4170.74 56
Pnet-blend78.13 4473.62 4581.13 4284.46 4288.20 3666.50 4580.74 4170.74 56
CasMVSNet(base)78.10 4668.51 5284.50 3685.47 3591.56 3259.79 5677.22 4776.47 44
Pnet-eth77.75 4785.52 3272.57 5775.57 5365.89 6181.44 3589.61 2976.26 45
MVSCRF77.21 4874.57 4478.98 4979.29 4983.87 4475.01 4374.12 5373.78 51
MVEpermissive76.82 1976.63 4977.52 4076.03 5470.40 5985.49 4275.68 4079.36 4372.20 53
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
MVSNet_plusplus76.50 5066.38 5383.25 3891.25 2873.79 5653.62 6279.14 4484.70 32
example74.48 5165.59 5480.41 4784.10 4783.16 4562.82 5068.36 5973.98 50
Snet74.39 5264.28 5781.13 4288.04 3373.18 5960.15 5568.41 5882.18 35
MVSNet73.65 5370.01 4976.07 5376.39 5277.59 5261.93 5278.09 4674.23 49
MVSNet + Gipuma72.61 5471.20 4773.54 5575.39 5473.46 5865.60 4776.81 5171.78 54
firsttry72.49 5575.56 4370.44 5968.80 6270.25 6069.31 4481.82 4072.28 52
MVSNet_++71.81 5659.04 6280.33 4884.86 3774.23 5533.06 6485.02 3681.88 37
test_112471.63 5763.97 5876.73 5169.39 6185.55 4156.73 5771.21 5575.26 46
CasMVSNet(SR_B)70.07 5871.18 4869.33 6069.73 6077.06 5364.06 4978.31 4561.20 61
F/T MVSNet+Gipuma69.80 5964.46 5673.36 5675.05 5573.57 5760.74 5368.18 6071.46 55
unMVSmet66.94 6060.67 6171.11 5871.91 5879.70 5054.62 5966.73 6161.74 60
confMetMVS63.51 6160.95 6065.21 6260.95 6376.48 5455.53 5866.38 6258.21 62
test_1120copyleft58.74 6257.03 6359.88 6349.42 6564.79 6249.65 6364.41 6365.45 59
PMVScopyleft70.75 2058.19 6343.22 6468.18 6173.06 5784.09 4354.23 6032.21 6447.37 63
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
Cas-MVS_preliminary56.71 6462.57 5952.80 6450.07 6461.96 6454.15 6171.00 5646.38 64
FADENet18.58 6515.51 6520.63 6521.45 6632.74 6615.09 6515.93 657.68 65
CMPMVSbinary69.68 2111.01 660.97 6617.70 6610.09 6743.01 651.93 660.00 660.00 66
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
dnet0.00 670.00 670.00 670.00 680.00 670.00 670.00 660.00 66
UnsupFinetunedMVSNet75.05 55