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-MVS93.56 186.37 181.38 289.69 291.68 290.34 179.19 283.57 387.05 2
PCF-MVS89.78 586.23 280.61 389.98 192.21 189.86 279.15 382.07 587.86 1
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
DeepC-MVS_fast93.52 284.79 378.95 688.68 390.85 388.97 476.51 681.39 886.21 5
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMP87.39 1084.24 478.65 887.96 488.90 588.89 574.85 1182.45 486.10 6
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
unMVSv184.05 583.50 184.41 1684.75 1887.18 1281.46 185.54 181.30 21
PLCcopyleft91.07 383.49 679.02 586.47 987.55 1187.15 1376.24 781.81 684.70 9
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
IB-MVS89.43 682.99 777.76 986.47 988.47 888.98 377.08 478.43 1481.96 19
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
ACMH+83.78 1582.73 875.83 1587.33 688.67 688.50 673.89 1277.77 1884.84 8
ACMM86.95 1382.58 976.68 1386.52 887.75 1087.28 1072.76 1580.59 1084.51 11
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
A-TVSNet + Gipumacopyleft82.55 1076.88 1186.32 1186.73 1387.81 972.89 1480.86 984.44 12
DeepC-MVS91.02 482.53 1174.98 1887.57 588.50 787.92 872.07 1977.88 1786.28 4
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
tm-dncc82.32 1280.42 483.58 1981.28 2782.48 2677.02 583.83 286.97 3
COLMAP(base)82.31 1376.85 1285.95 1287.50 1286.55 1575.41 1078.29 1583.82 13
COLMAP(SR)81.85 1473.79 1987.23 789.02 487.99 772.66 1774.92 1984.68 10
A1Net81.22 1575.74 1684.87 1485.61 1686.60 1472.67 1678.82 1282.39 17
BP-MVSNet81.15 1675.41 1784.97 1387.81 981.23 3172.11 1878.70 1385.88 7
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
GSE80.93 1778.72 782.41 2282.36 2481.72 3075.91 881.53 783.15 15
TAPA-MVS87.50 980.68 1875.95 1483.83 1884.25 2184.68 1973.62 1378.28 1682.57 16
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
ACMH83.09 1779.47 1971.54 2084.76 1586.56 1485.52 1770.21 2172.87 2282.19 18
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
TAPA-MVS(SR)78.74 2070.70 2384.11 1784.55 1984.37 2067.20 2474.20 2083.39 14
COLMAP_ROBcopyleft82.69 1878.27 2171.49 2182.79 2083.46 2283.41 2469.96 2273.03 2181.51 20
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
LTVRE_ROB81.71 1974.79 2271.44 2277.02 3178.37 3073.56 4370.22 2072.67 2379.12 22
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
3Dnovator+87.72 874.01 2362.76 2781.51 2385.71 1585.55 1661.51 2664.01 3073.28 27
PVSNet_083.28 1673.62 2460.09 3182.64 2185.11 1787.22 1159.21 2960.96 3475.59 24
3Dnovator87.35 1173.38 2562.96 2680.33 2484.34 2084.16 2261.54 2564.39 2972.49 30
HY-MVS88.56 772.41 2662.56 2878.98 2779.20 2983.22 2558.72 3166.40 2774.52 25
OpenMVScopyleft85.28 1472.17 2761.03 3079.59 2682.41 2383.50 2358.91 3063.15 3272.86 29
PVSNet87.13 1271.39 2858.61 3279.90 2581.65 2584.27 2159.32 2857.90 3773.79 26
CPR_FA71.34 2961.72 2977.76 2976.15 3379.96 3460.48 2762.95 3377.16 23
LPCS70.58 3067.95 2472.33 3470.49 3973.62 4267.73 2368.16 2572.89 28
ANet-0.7569.45 3177.48 1064.10 4661.68 5169.60 4975.76 979.19 1161.02 41
tmmvs67.95 3253.69 3777.45 3081.28 2782.48 2653.71 3453.66 4368.60 32
CIDER67.82 3351.67 3978.59 2881.40 2684.83 1852.81 3550.54 4769.54 31
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
test_mvsss65.81 3455.91 3472.41 3374.70 3680.45 3245.55 4266.28 2862.09 40
test_120565.70 3556.79 3371.64 3774.86 3581.89 2949.86 3963.72 3158.17 45
OpenMVS_ROBcopyleft73.86 2063.94 3650.71 4172.76 3273.57 3779.97 3350.36 3651.06 4664.76 36
R-MVSNet63.29 3755.31 3568.61 4066.45 4474.27 4055.32 3355.31 4165.10 35
PVSNet_LR62.94 3849.73 4271.75 3671.11 3876.61 3843.79 4455.66 4067.54 33
PSD-MVSNet60.65 3954.96 3664.44 4465.36 4665.69 5549.87 3860.04 3562.27 39
test_112659.75 4043.11 4970.84 3876.19 3277.23 3741.68 4744.53 5459.11 43
P-MVSNet58.87 4163.50 2555.79 5651.35 6559.52 6258.47 3268.53 2456.50 48
SGNet56.81 4250.90 4060.74 5162.11 5061.27 5944.65 4357.16 3858.84 44
MVS_test_156.76 4345.41 4664.33 4567.23 4378.33 3539.26 5051.56 4447.43 57
MVSNet_++55.13 4435.61 5668.15 4175.11 3468.50 5311.93 8359.29 3660.82 42
test354.74 4547.91 4359.30 5360.21 5361.13 6041.75 4654.07 4256.54 47
unsupervisedMVS_cas54.14 4640.65 5263.13 4763.75 4869.37 5133.91 5747.39 4956.26 49
ANet54.11 4743.47 4761.21 4961.68 5169.60 4941.77 4545.17 5352.35 53
Snet53.94 4826.41 7272.29 3569.83 4081.92 2819.51 7433.31 6865.12 34
AttMVS51.70 4946.70 4455.04 5948.03 6762.78 5747.24 4046.15 5154.32 51
CasMVSNet(SR_A)51.60 5032.28 6364.48 4366.36 4574.08 4125.77 6738.79 6053.01 52
Pnet_fast51.22 5123.37 7569.79 3967.40 4278.24 3615.90 7630.84 7263.72 37
Pnet-new-50.72 5227.58 6866.15 4264.59 4771.56 4427.24 6227.92 7462.31 38
mvs_zhu_103050.59 5337.13 5559.57 5257.86 5966.44 5430.92 5943.34 5554.39 50
TVSNet49.80 5442.25 5054.83 6057.91 5856.82 6534.18 5550.32 4849.77 55
CasMVSNet(base)49.64 5530.94 6562.11 4863.47 4971.04 4724.30 7137.57 6351.83 54
MVEpermissive44.00 2249.53 5641.59 5154.82 6149.03 6669.81 4836.17 5447.01 5045.62 59
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
MVSNet_plusplus48.34 5729.36 6660.99 5069.62 4156.31 6620.92 7237.80 6157.05 46
firsttry48.32 5839.93 5453.90 6253.09 6363.57 5634.00 5645.86 5245.05 60
RMVSNet47.98 5952.85 3844.74 7055.96 6047.68 7439.09 5166.61 2630.56 73
example47.27 6033.53 5956.43 5559.94 5575.51 3931.42 5835.63 6433.83 70
hgnet46.66 6133.97 5755.13 5759.48 5671.40 4536.55 5231.38 6934.50 68
DPSNet46.66 6133.97 5755.13 5759.48 5671.40 4536.55 5231.38 6934.50 68
metmvs_fine46.00 6345.56 4546.29 6953.05 6445.70 7639.94 4951.18 4540.12 64
QQQNet45.34 6440.26 5348.73 6660.02 5458.55 6339.99 4840.53 5627.62 76
MVSNet40.88 6532.15 6446.71 6840.86 7250.77 7326.51 6437.78 6248.50 56
MVSCRF40.84 6626.49 7150.40 6347.42 7058.10 6427.02 6325.96 7545.69 58
Pnet-blend++39.83 6724.24 7350.22 6454.22 6155.24 6714.25 7934.23 6541.21 62
Pnet-blend39.83 6724.24 7350.22 6454.22 6155.24 6714.25 7934.23 6541.21 62
MVSNet + Gipuma37.61 6928.66 6743.58 7239.30 7353.28 7026.38 6530.95 7138.16 65
SVVNet37.34 7032.93 6040.27 7447.52 6844.02 7825.34 6940.53 5629.29 74
ternet37.34 7032.93 6040.27 7447.52 6844.02 7825.34 6940.53 5629.29 74
Pnet-eth37.00 7243.12 4832.92 7939.11 7521.96 8430.33 6055.90 3937.68 66
F/T MVSNet+Gipuma36.65 7326.87 6943.17 7338.81 7653.12 7125.52 6828.22 7337.56 67
test_112436.46 7419.17 7747.98 6741.35 7160.36 6115.94 7522.39 7642.24 61
CCVNet34.56 7526.72 7039.79 7639.28 7452.47 7219.64 7333.80 6727.62 76
CasMVSNet(SR_B)34.04 7632.49 6235.07 7832.66 8045.95 7526.11 6638.87 5926.61 78
vp_mvsnet33.19 7716.83 7844.10 7137.77 7962.70 5814.98 7818.68 7731.84 71
PMVScopyleft41.42 2331.23 7820.08 7638.66 7738.19 7853.59 6928.97 6111.19 8224.20 79
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
unMVSmet26.14 7916.71 7932.43 8029.12 8144.88 7715.84 7717.58 7823.29 80
confMetMVS22.01 8015.55 8026.32 8121.89 8238.61 8113.95 8117.14 7918.45 82
Cas-MVS_preliminary19.83 8113.19 8124.25 8217.25 8339.23 8012.65 8213.73 8116.28 83
test_1120copyleft16.76 8212.65 8219.50 8311.72 8423.76 8310.31 8414.98 8023.04 81
FADENet5.54 833.06 837.20 853.62 8516.50 853.91 852.21 831.49 84
CMPMVSbinary58.40 215.41 840.12 848.94 842.23 8624.58 820.25 860.00 840.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
dnet0.00 850.00 850.00 860.00 870.00 860.00 870.00 840.00 85
test_MVS50.36 36
test_robustmvs58.94 5477.56 3168.56 5246.73 4130.70 72
UnsupFinetunedMVSNet38.81 76