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 bysort bysort bysort bysorted bysort bysort bysort bysort by
CasMVSNet(SR_B)65.31 944.19 2779.39 191.10 174.56 435.00 3353.38 2772.51 7
AttMVS64.84 1057.49 869.74 1388.51 250.94 3445.77 1569.21 569.78 12
MVSNet65.82 852.08 1674.99 488.07 385.21 141.08 2063.08 751.67 38
3Dnovator80.37 769.73 165.51 372.55 1082.32 465.67 1557.53 373.48 169.67 13
3Dnovator+83.92 269.39 362.05 574.29 781.76 564.89 1752.69 771.42 376.22 1
LTVRE_ROB86.10 169.45 258.19 776.95 281.50 678.07 355.83 560.55 1471.29 9
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
tm-dncc63.90 1248.20 2374.37 681.20 766.01 1338.20 2458.21 1875.91 2
tmmvs66.88 456.70 1073.68 981.20 766.01 1350.58 862.81 1073.82 6
PMVScopyleft80.48 657.42 2129.11 4876.30 379.61 979.41 234.22 3524.00 5469.88 11
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
test_1120copyleft63.22 1368.85 159.47 3178.81 1051.99 3379.45 158.25 1747.61 46
DeepPCF-MVS81.24 566.76 554.74 1374.77 576.18 1173.22 546.42 1463.06 974.91 3
DeepC-MVS_fast80.27 865.96 654.23 1473.79 875.88 1271.59 645.37 1763.08 773.89 5
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
mvs_zhu_103062.39 1450.41 2070.37 1275.70 1366.39 1240.10 2260.73 1369.02 14
OpenMVScopyleft76.72 1364.07 1159.05 667.42 1575.57 1462.58 1947.98 1070.11 464.12 19
DeepC-MVS82.31 465.89 757.22 971.66 1174.01 1570.18 747.73 1166.72 670.79 10
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
PCF-MVS74.62 1556.56 2343.50 3065.27 1773.58 1659.98 2437.23 2749.76 3362.24 22
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
test_112461.69 1667.72 257.67 3471.65 1746.06 4275.31 260.13 1555.30 33
test_112658.81 1854.80 1161.48 2170.94 1856.27 2753.41 656.19 2157.23 31
GSE53.18 3140.31 3661.77 2069.15 1952.25 3235.20 3245.43 3963.90 20
Pnet-new-62.28 1564.63 460.72 2368.45 2045.71 4356.04 473.22 267.99 15
LPCS53.86 2942.82 3161.22 2267.38 2154.15 2837.38 2648.27 3662.14 23
TAPA-MVS(SR)57.16 2254.76 1258.76 3267.29 2248.19 3850.15 959.37 1660.81 24
P-MVSNet54.95 2750.02 2258.24 3366.31 2347.74 4041.88 1958.16 1960.67 25
OpenMVS_ROBcopyleft70.19 1751.98 3450.29 2153.10 3565.07 2439.38 4842.82 1857.77 2054.85 34
COLMAP_ROBcopyleft83.01 361.51 1750.83 1968.63 1465.07 2466.61 940.64 2161.02 1274.20 4
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
TAPA-MVS77.73 1258.58 1953.61 1561.89 1964.81 2660.73 2145.40 1661.82 1160.14 28
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
CasMVSNet(base)55.44 2643.59 2963.33 1863.97 2765.66 1632.47 3754.72 2460.37 27
COLMAP(base)57.56 2045.74 2465.45 1663.08 2861.37 2039.37 2352.10 3071.91 8
Pnet-blend++56.52 2451.51 1759.87 2662.39 2966.50 1047.44 1255.57 2250.71 40
Pnet-blend56.52 2451.51 1759.87 2662.39 2966.50 1047.44 1255.57 2250.71 40
COLMAP(SR)52.61 3241.79 3359.83 2858.75 3154.00 2937.15 2846.43 3866.74 16
ACMM79.39 952.37 3340.99 3559.96 2558.56 3260.72 2231.00 3850.98 3160.62 26
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
CasMVSNet(SR_A)53.46 3044.06 2859.72 2957.37 3362.68 1834.23 3453.90 2659.09 30
ACMP79.16 1051.84 3540.13 3859.65 3057.20 3458.68 2530.46 3949.80 3263.09 21
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
ANet-0.7536.48 4914.29 6551.28 3755.77 3552.26 3010.99 6317.58 6445.82 47
ANet43.24 4229.49 4652.40 3655.77 3552.26 3027.30 4331.68 4849.19 43
PLCcopyleft73.85 1654.06 2845.01 2560.08 2455.47 3760.30 2337.58 2552.44 2964.49 17
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMH+77.89 1147.08 3642.29 3250.27 4054.95 3839.16 4935.24 3049.34 3456.70 32
MVSNet_plusplus36.25 5026.59 5042.70 4754.00 3914.25 705.76 8047.42 3759.84 29
ACMH76.49 1445.64 3841.63 3448.31 4353.97 4038.96 5034.00 3649.25 3552.00 36
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
vp_mvsnet42.75 4330.63 4450.83 3853.01 4147.99 398.57 7352.69 2851.49 39
HY-MVS64.64 1844.49 3940.14 3747.40 4452.03 4242.24 4636.35 2943.93 4147.92 45
CIDER46.95 3744.89 2648.33 4250.37 4346.20 4135.21 3154.58 2548.41 44
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
PVSNet_LR43.87 4133.54 4250.75 3949.15 4450.81 3523.32 4743.77 4252.30 35
Pnet_fast44.24 4035.70 3949.94 4146.98 4558.15 2629.16 4142.24 4444.70 48
IB-MVS62.13 1939.31 4435.63 4041.76 5046.97 4638.95 5126.86 4444.41 4039.36 50
Christian Sormann, Mattia Rossi, Andreas Kuhn and Friedrich Fraundorfer: IB-MVS: An Iterative Algorithm for Deep Multi-View Stereo based on Binary Decisions. BMVC 2021
PVSNet_051.08 2239.05 4529.60 4545.34 4544.29 4739.84 4726.44 4532.77 4751.89 37
PVSNet58.17 2138.39 4631.41 4343.04 4641.89 4837.34 5229.69 4033.13 4649.88 42
R-MVSNet37.45 4734.59 4139.35 5141.64 4942.85 4525.44 4643.75 4333.57 54
test_120536.84 4829.16 4741.97 4941.52 5049.04 3627.34 4230.98 5035.34 53
MVSCRF34.84 5129.04 4938.71 5241.35 5144.92 4418.79 5139.30 4529.86 55
BP-MVSNet32.65 5323.64 5138.65 5340.81 5233.99 5419.28 4928.00 5141.15 49
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
CMPMVSbinary59.41 2021.70 550.34 8335.93 5438.78 5369.01 80.69 850.00 840.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
A-TVSNet + Gipumacopyleft33.17 5219.66 5442.17 4838.64 5448.93 3720.23 4819.09 6138.94 51
MVS_test_121.23 5616.85 5824.15 5731.31 5518.73 616.46 7927.23 5222.39 61
unsupervisedMVS_cas26.15 5418.92 5530.97 5530.39 5634.23 5315.24 5422.60 5628.30 56
Pnet-eth20.12 5918.76 5721.02 6024.61 5721.44 5713.26 5624.26 5317.02 67
MVSNet_++21.10 5716.58 6124.11 5822.79 5810.88 761.86 8431.30 4938.65 52
Snet19.16 6019.74 5218.78 6122.45 5915.11 6716.74 5322.73 5518.76 65
example11.91 757.31 8114.98 6820.88 6020.72 588.09 756.53 823.34 82
hgnet14.22 7013.23 6614.89 6918.80 6120.68 5910.37 6816.09 655.18 80
DPSNet14.22 7013.23 6614.89 6918.80 6120.68 5910.37 6816.09 655.18 80
F/T MVSNet+Gipuma18.31 6119.68 5317.39 6317.36 6313.22 7218.92 5020.43 5821.60 62
UnsupFinetunedMVSNet17.36 63
CCVNet17.35 6311.44 7121.29 5916.96 6521.70 5612.75 5910.13 7925.21 57
MVSNet + Gipuma17.51 6218.78 5616.66 6516.50 6612.75 7318.63 5218.92 6220.74 63
test_mvsss14.34 698.77 7918.06 6216.23 6723.38 554.71 8112.84 7114.55 70
Cas-MVS_preliminary20.88 5811.16 7427.36 5612.72 685.17 8111.58 6210.75 7864.20 18
A1Net14.88 6515.39 6214.54 7312.69 6911.35 7512.32 6018.45 6319.59 64
firsttry10.67 798.36 8012.20 7612.25 7010.50 777.87 768.86 8013.86 71
CPR_FA15.63 6416.76 5914.87 7112.15 7114.63 6913.13 5720.39 5917.84 66
TVSNet13.85 7212.75 6914.58 7211.20 7215.62 6510.72 6514.79 6816.93 68
test313.52 7312.47 7014.23 7410.80 7315.75 6210.84 6414.10 6916.14 69
QQQNet14.60 6611.15 7516.90 6410.35 7415.13 669.88 7012.42 7325.21 57
SVVNet14.42 6711.43 7216.41 6610.16 7515.66 6310.44 6612.42 7323.41 59
ternet14.42 6711.43 7216.41 6610.16 7515.66 6310.44 6612.42 7323.41 59
SGNet11.80 7610.60 7612.60 759.98 7714.69 688.98 7112.23 7613.13 72
PSD-MVSNet11.08 789.72 7812.00 779.76 7813.96 718.14 7411.29 7712.27 73
unMVSv17.70 836.95 828.20 808.59 799.59 786.96 776.95 816.43 77
unMVSmet11.69 7716.62 608.41 798.35 808.22 7912.26 6120.99 578.66 76
MVEpermissive40.22 2312.38 7414.61 6310.89 787.98 8112.62 7413.54 5515.69 6712.06 74
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
confMetMVS8.11 829.85 776.95 817.36 824.72 836.48 7813.22 708.77 75
RMVSNet9.14 8012.82 686.69 827.06 837.73 8012.89 5812.74 725.29 79
metmvs_fine9.11 8114.53 645.49 835.49 844.77 828.81 7220.25 606.22 78
test_robustmvs1.67 842.86 851.42 842.72 830.73 83
FADENet0.15 840.18 840.13 850.22 860.09 850.23 860.12 830.07 84
dnet0.00 850.00 850.00 860.00 870.00 860.00 870.00 840.00 85
test_MVS4.45 82