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
DeepPCF-MVS93.97 187.63 182.10 191.31 191.80 191.34 177.64 186.56 190.79 1
DeepC-MVS_fast93.89 286.59 280.39 290.73 291.00 390.77 275.90 384.89 390.41 2
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepC-MVS93.07 385.88 379.73 489.98 389.82 690.09 574.93 584.53 490.03 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
TAPA-MVS90.10 784.80 480.13 387.91 688.22 888.54 975.35 484.91 286.96 14
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
PCF-MVS89.48 1084.51 576.96 1089.54 491.30 288.76 773.35 880.56 1288.57 4
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
TAPA-MVS(SR)83.88 678.08 787.74 888.99 786.49 1673.17 1082.99 687.76 10
COLMAP(base)83.59 777.50 987.65 986.74 1287.68 1173.30 981.70 988.52 5
PLCcopyleft91.00 683.54 878.38 586.98 1385.38 1687.33 1373.45 683.31 588.23 9
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
GSE83.39 977.82 887.10 1285.70 1587.27 1473.41 782.23 788.34 8
COLMAP(SR)83.28 1076.24 1287.97 587.61 987.90 1072.76 1179.71 1488.39 7
ACMH+87.92 1382.46 1176.63 1186.35 1686.95 1184.52 1871.68 1281.58 1087.56 12
ACMP89.59 982.31 1275.00 1487.18 1184.49 2088.57 868.43 1881.58 1088.49 6
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
LTVRE_ROB88.41 1282.23 1378.11 684.97 1986.24 1382.83 2476.93 279.30 1585.84 16
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
ACMM89.79 882.20 1475.34 1386.76 1484.94 1887.68 1168.85 1781.84 887.67 11
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
3Dnovator91.36 582.19 1574.09 1687.59 1089.83 590.10 471.54 1376.63 1882.84 20
3Dnovator+91.43 481.91 1673.11 1987.78 790.06 490.17 370.40 1575.82 2183.12 18
COLMAP_ROBcopyleft87.81 1481.57 1774.74 1586.13 1783.81 2187.08 1569.46 1680.01 1387.49 13
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
ACMH87.59 1580.55 1873.69 1785.11 1886.17 1482.81 2568.19 1979.20 1686.36 15
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
OpenMVScopyleft89.19 1180.38 1971.40 2086.37 1587.42 1089.29 667.70 2075.10 2282.40 22
LPCS78.54 2073.62 1881.82 2279.59 2783.19 2271.25 1475.99 2082.69 21
A-TVSNet + Gipumacopyleft77.14 2167.96 2383.26 2080.53 2486.26 1765.25 2170.67 2382.99 19
BP-MVSNet76.76 2268.58 2282.22 2184.51 1977.66 3460.98 2676.18 1984.48 17
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
PVSNet_082.17 1774.27 2364.50 2580.78 2381.77 2380.64 2861.11 2467.89 2579.92 23
PVSNet86.66 1673.29 2465.21 2478.67 2780.40 2577.98 3264.04 2366.38 2877.62 24
OpenMVS_ROBcopyleft81.14 1873.28 2563.11 2780.06 2582.70 2281.62 2761.11 2465.12 3075.87 25
CIDER72.92 2662.80 2879.66 2680.16 2683.22 2160.69 2764.90 3175.61 27
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
test_112671.43 2757.57 3280.67 2485.16 1784.45 1955.16 3059.98 3572.40 29
R-MVSNet70.54 2863.98 2674.91 3173.85 3377.87 3359.85 2868.11 2473.01 28
P-MVSNet70.42 2971.20 2169.90 3566.73 3872.02 3664.84 2277.55 1770.95 31
ANet-0.7567.51 3058.66 3173.41 3370.12 3479.37 2950.72 3266.61 2770.74 32
AttMVS65.89 3160.14 2969.73 3664.68 4175.04 3558.80 2961.47 3469.47 34
ANet63.85 3252.65 3471.32 3470.12 3479.37 2948.53 3456.77 3864.48 39
Pnet-new-63.42 3342.13 4477.61 2877.71 2979.30 3141.06 3743.20 5075.82 26
CasMVSNet(SR_A)63.08 3445.33 4174.92 3074.36 3083.73 2037.04 4453.62 4166.66 37
CasMVSNet(base)61.88 3543.72 4273.99 3274.03 3281.95 2634.99 4752.45 4265.99 38
Pnet_fast59.94 3635.26 5676.39 2974.15 3182.99 2325.02 5545.51 4772.03 30
CPR_FA58.26 3756.35 3359.54 4456.43 4855.34 4750.61 3362.09 3366.83 36
MVSNet56.22 3846.49 3862.70 4057.79 4567.16 4238.93 4054.05 3963.16 41
A1Net54.85 3959.98 3051.43 4747.46 5742.97 5652.72 3167.24 2663.86 40
MVSNet_plusplus54.10 4036.12 5466.09 3778.36 2849.55 5120.40 6051.84 4370.36 33
Pnet-blend++53.68 4137.04 5164.76 3868.91 3670.02 3824.13 5649.96 4555.36 45
Pnet-blend53.68 4137.04 5164.76 3868.91 3670.02 3824.13 5649.96 4555.36 45
MVSCRF51.99 4337.64 4861.55 4258.88 4268.35 4034.99 4740.28 5357.43 44
MVSNet_++50.85 4437.40 4959.81 4365.34 4046.28 529.19 6365.60 2967.81 35
test_112450.28 4531.72 5762.65 4158.29 4470.82 3727.23 5436.21 5758.84 42
Snet49.32 4636.02 5558.18 4566.02 3950.24 4927.93 5344.12 4858.27 43
CasMVSNet(SR_B)48.99 4745.75 3951.15 4849.07 5562.78 4337.57 4153.93 4041.61 52
Pnet-eth45.62 4849.15 3643.27 5753.31 5029.96 6435.03 4663.26 3246.53 49
unMVSv145.02 4943.12 4346.29 5447.83 5650.14 5042.67 3543.57 4940.92 53
RMVSNet44.67 5050.67 3540.67 5850.79 5141.21 5842.32 3659.02 3630.02 57
MVEpermissive50.73 2144.67 5045.57 4044.07 5635.90 6052.30 4840.89 3850.26 4444.00 51
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
DPSNet44.61 5238.48 4648.69 4957.33 4660.08 4437.17 4239.80 5528.65 59
hgnet44.61 5238.48 4648.69 4957.33 4660.08 4437.17 4239.80 5528.65 59
MVSNet + Gipuma44.11 5438.76 4547.67 5250.05 5445.26 5434.93 4942.60 5147.71 47
F/T MVSNet+Gipuma43.59 5537.31 5047.78 5150.12 5245.64 5334.25 5040.37 5247.60 48
PMVScopyleft53.92 2042.74 5626.50 5953.57 4654.03 4968.05 4136.35 4516.65 6338.64 54
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
firsttry41.11 5736.23 5344.36 5543.82 5843.77 5532.28 5140.18 5445.51 50
example40.49 5830.54 5847.11 5358.75 4358.88 4629.97 5231.12 5823.71 63
metmvs_fine39.41 5948.46 3733.38 6035.53 6131.67 6339.28 3957.64 3732.95 56
unMVSmet32.09 6025.89 6036.22 5936.96 5942.02 5723.79 5827.99 5929.68 58
test_1120copyleft27.94 6122.22 6231.76 6120.92 6437.22 6018.54 6225.90 6137.14 55
confMetMVS26.62 6223.72 6128.55 6328.79 6233.00 6221.33 5926.10 6023.86 62
Cas-MVS_preliminary25.33 6319.98 6328.90 6224.65 6334.26 6118.62 6121.35 6227.78 61
CMPMVSbinary62.92 198.44 640.22 6513.92 644.26 6537.50 590.45 650.00 650.00 65
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
FADENet1.48 651.79 641.27 652.11 661.06 652.23 641.34 640.65 64
dnet0.00 660.00 660.00 660.00 670.00 660.00 660.00 650.00 65
UnsupFinetunedMVSNet50.12 52