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
IB-MVS77.80 467.29 153.24 676.65 177.55 178.10 153.88 152.59 974.31 1
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
BP-MVSNet64.34 250.31 973.69 274.99 271.90 1146.42 1254.19 874.18 2
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
PVSNet_068.08 1559.14 1042.70 2070.10 374.33 374.52 338.12 2247.28 1561.46 15
PVSNet73.49 857.76 1643.28 1967.41 1070.29 472.10 940.29 2046.26 1859.84 17
COLMAP(SR)60.84 647.20 1369.93 469.97 572.99 647.93 746.47 1666.85 4
A1Net62.59 354.91 367.70 968.95 672.10 952.15 357.67 362.06 13
HY-MVS76.49 557.81 1540.66 2269.25 668.59 773.61 535.90 2845.41 2065.54 8
TAPA-MVS(SR)55.57 1735.84 3168.73 867.52 872.24 833.19 3138.50 3066.42 6
ACMP71.68 1062.33 451.03 869.86 566.41 974.93 247.36 854.70 768.24 3
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
ACMH+65.35 1659.83 846.18 1668.93 765.70 1074.24 446.34 1346.01 1966.83 5
DeepC-MVS_fast79.48 259.14 1048.44 1166.27 1164.81 1170.57 1346.73 1150.15 1263.44 10
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
A-TVSNet + Gipumacopyleft61.02 556.34 164.14 1564.32 1266.91 2452.09 460.60 161.19 16
DeepPCF-MVS81.17 160.64 753.51 565.40 1464.27 1369.40 1851.28 655.74 662.53 12
DeepC-MVS77.85 359.42 949.21 1066.23 1364.01 1470.42 1546.20 1552.23 1064.26 9
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
MVSNet_++41.03 4219.38 5555.47 3363.04 1557.24 406.23 7732.52 4046.14 40
TAPA-MVS70.22 1258.89 1251.21 764.01 1662.59 1666.45 2646.31 1456.11 562.99 11
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
3Dnovator+73.60 753.20 2138.24 2663.18 1862.18 1770.72 1236.81 2339.68 2856.64 23
CIDER49.33 2733.81 3459.69 2661.47 1868.26 1931.50 3336.11 3649.34 33
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
PCF-MVS73.15 958.42 1354.11 461.29 2061.25 1964.94 3051.80 556.43 457.66 21
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
3Dnovator73.91 652.81 2238.17 2762.58 1960.95 2069.63 1735.98 2740.36 2757.16 22
ACMM69.62 1358.27 1446.31 1566.24 1260.67 2172.46 742.63 1850.00 1365.60 7
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
Snet41.99 4016.05 6359.29 2860.27 2270.55 149.62 7222.48 6147.04 38
ACMH63.93 1752.68 2435.79 3263.93 1760.12 2370.07 1634.99 2936.59 3561.61 14
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
OpenMVS_ROBcopyleft61.12 1848.48 2831.63 3559.71 2559.03 2467.15 2231.08 3432.18 4152.95 29
COLMAP(base)54.01 1944.78 1760.17 2258.77 2563.60 3245.57 1643.99 2258.14 20
PVSNet_LR46.69 3227.25 4159.64 2758.01 2667.25 2120.54 4633.97 3853.67 28
PLCcopyleft68.80 1453.38 2047.19 1457.51 2957.17 2761.22 3646.82 1047.56 1454.14 25
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
test_120545.54 3530.23 3755.74 3256.56 2866.59 2521.99 4438.47 3144.08 44
OpenMVScopyleft70.45 1151.01 2637.34 2860.13 2356.52 2967.91 2033.72 3040.96 2655.96 24
PSD-MVSNet47.65 3039.40 2453.15 3756.48 3053.80 4732.51 3246.29 1749.16 34
tm-dncc55.48 1848.30 1260.27 2156.02 3165.28 2844.96 1751.65 1159.50 18
tmmvs46.16 3330.02 3856.92 3056.02 3165.28 2830.14 3529.91 4549.47 32
GSE52.65 2541.64 2159.98 2455.71 3365.76 2741.60 1941.69 2558.48 19
test_112640.23 4319.47 5454.08 3655.58 3462.13 3515.66 5423.27 5644.52 43
CasMVSNet(SR_A)41.31 4118.44 5956.56 3155.29 3567.08 2313.89 6222.98 5847.30 37
test_robustmvs33.07 6153.99 3631.42 7526.47 3913.81 72
SGNet44.07 3634.19 3350.65 4353.93 3751.56 5526.54 3841.85 2446.47 39
Pnet_fast36.12 5212.02 7052.20 3952.81 3863.56 336.47 7617.57 6540.22 47
test342.75 3931.34 3650.36 4452.00 3953.28 4923.88 4138.80 2945.79 41
QQQNet34.46 5323.92 4541.49 5751.88 4050.62 5722.82 4325.03 5121.97 66
LPCS48.13 2938.62 2554.48 3551.49 4157.98 3938.93 2138.30 3353.97 26
Pnet-new-38.13 4713.54 6654.53 3450.63 4263.16 3415.34 5711.75 7149.79 30
CasMVSNet(base)38.80 4417.46 6153.03 3850.40 4363.89 3113.07 6421.85 6244.80 42
COLMAP_ROBcopyleft57.96 2045.89 3437.03 3051.79 4049.43 4456.42 4536.78 2437.28 3449.53 31
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
TVSNet38.58 4526.99 4246.30 4849.23 4549.73 5819.13 4734.85 3739.94 48
test_mvsss33.68 5520.19 5142.68 5247.39 4651.80 5317.54 5122.84 5928.85 58
CPR_FA47.15 3140.41 2351.65 4147.19 4753.80 4736.70 2544.12 2153.96 27
MVS_test_134.36 5422.83 4642.05 5546.73 4855.32 4617.93 4927.73 4924.10 63
MVSNet_plusplus29.36 589.54 7342.57 5346.21 4944.17 636.65 7512.43 7037.33 53
LTVRE_ROB59.60 1943.60 3737.31 2947.80 4645.94 5049.52 5936.19 2638.43 3247.93 36
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
ANet-0.7552.71 2355.63 250.76 4244.38 5159.75 3753.05 258.21 248.17 35
ANet37.00 5121.13 4947.57 4744.38 5159.75 3718.53 4823.72 5538.59 50
unsupervisedMVS_cas37.45 4821.88 4847.84 4543.96 5356.86 4115.49 5528.27 4842.70 45
R-MVSNet37.16 4926.49 4344.27 4942.37 5452.02 5124.06 4028.93 4738.41 52
unMVSv143.21 3844.62 1842.27 5440.66 5551.71 5447.08 942.16 2334.44 54
mvs_zhu_103033.65 5618.58 5843.70 5139.88 5651.94 5214.13 5923.03 5739.29 49
P-MVSNet38.28 4629.70 4044.00 5039.20 5750.96 5629.12 3630.29 4441.85 46
Pnet-blend++23.49 679.16 7433.04 6238.23 5835.75 723.77 8114.54 6825.15 60
Pnet-blend23.49 679.16 7433.04 6238.23 5835.75 723.77 8114.54 6825.15 60
SVVNet27.23 6019.50 5232.39 6535.04 6038.67 6813.98 6025.03 5123.46 64
ternet27.23 6019.50 5232.39 6535.04 6038.67 6813.98 6025.03 5123.46 64
AttMVS37.07 5030.02 3841.77 5634.23 6252.58 5028.11 3731.93 4238.51 51
firsttry32.39 5725.47 4436.99 5932.51 6346.62 6121.58 4529.36 4631.86 56
test_112425.42 637.36 7637.46 5831.97 6449.24 608.21 736.51 7631.19 57
example24.50 6516.56 6229.80 6827.47 6556.53 4417.69 5015.43 665.40 80
MVSCRF24.97 6413.50 6732.61 6426.31 6643.85 6412.02 6614.98 6727.67 59
MVSNet27.67 5918.22 6033.96 6026.28 6743.46 6513.67 6322.78 6032.14 55
CCVNet23.74 6614.42 6429.95 6725.90 6841.97 679.71 7119.13 6321.97 66
hgnet22.90 6913.03 6829.49 7025.13 6956.72 4216.70 529.35 746.62 78
DPSNet22.90 6913.03 6829.49 7025.13 6956.72 4216.70 529.35 746.62 78
vp_mvsnet19.19 744.47 7829.00 7222.46 7145.93 625.19 783.74 7818.61 70
MVEpermissive24.84 2326.04 6220.59 5029.66 6922.29 7242.00 6615.49 5525.70 5024.70 62
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
CasMVSNet(SR_B)21.94 7119.24 5623.74 7521.33 7334.30 7414.54 5823.93 5415.57 71
Pnet-eth14.99 7618.95 5712.35 7719.54 744.23 847.59 7430.32 4313.27 73
F/T MVSNet+Gipuma19.68 7211.34 7125.23 7316.34 7538.41 7011.15 6811.54 7220.95 68
UnsupFinetunedMVSNet16.34 75
MVSNet + Gipuma19.59 7311.32 7225.11 7416.29 7738.21 7111.18 6711.45 7320.83 69
PMVScopyleft26.43 2213.46 777.26 7717.59 7614.19 7830.85 7610.42 694.11 777.73 77
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
RMVSNet15.61 7522.67 4710.90 7813.04 7914.34 7912.27 6533.07 395.30 81
metmvs_fine12.14 7814.21 6510.76 798.78 8013.01 8110.35 7018.07 6410.49 74
unMVSmet7.50 793.60 7910.10 806.98 8118.02 783.83 803.37 795.28 82
Cas-MVS_preliminary6.17 802.14 818.86 814.86 8213.44 803.08 831.20 828.27 76
test_1120copyleft5.30 813.59 806.45 844.63 834.96 833.98 793.19 809.76 75
confMetMVS4.71 832.07 826.46 834.51 8411.02 821.90 842.24 813.85 83
CMPMVSbinary48.56 214.84 820.02 848.06 821.22 8522.96 770.03 860.00 840.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
FADENet0.32 840.12 830.45 850.50 860.77 850.16 850.09 830.08 84
dnet0.00 850.00 850.00 860.00 870.00 860.00 870.00 840.00 85
test_MVS23.36 42