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-MVS69.38 247.34 138.25 153.40 254.29 456.62 332.95 143.55 249.30 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepPCF-MVS69.58 146.90 237.52 353.15 353.58 557.05 232.83 242.22 348.83 5
DeepC-MVS_fast68.24 346.61 334.94 454.40 155.37 357.55 131.02 438.86 450.28 1
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
3Dnovator64.47 545.24 434.37 552.48 557.34 153.53 530.63 538.11 646.56 8
3Dnovator+66.72 444.84 533.14 752.63 456.23 253.78 429.20 637.07 947.89 6
ACMP63.53 642.50 631.11 950.08 648.62 1052.78 724.90 1537.33 848.84 4
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
OpenMVScopyleft61.03 942.28 732.86 848.56 951.94 750.82 1027.02 938.70 542.92 12
TAPA-MVS59.36 1041.97 837.54 244.93 1346.78 1446.52 1631.44 343.65 141.48 15
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
COLMAP(SR)41.20 929.58 1048.95 748.70 948.18 1228.26 830.89 1749.98 2
ACMM61.98 740.80 1029.05 1248.64 846.87 1253.22 623.47 1934.63 1045.82 9
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
TAPA-MVS(SR)40.72 1129.34 1148.31 1052.32 644.77 2026.63 1132.05 1347.84 7
PCF-MVS61.88 840.12 1233.16 644.75 1449.39 846.16 1728.63 737.70 738.70 20
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
COLMAP(base)37.35 1328.09 1843.52 1642.34 2146.68 1526.55 1229.63 2041.53 14
LTVRE_ROB55.42 1536.88 1428.57 1542.42 1840.72 2548.22 1126.97 1030.17 1938.32 21
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
ACMH+57.40 1136.48 1528.86 1341.57 2043.89 1937.61 3326.39 1331.33 1543.20 11
CasMVSNet(SR_A)35.87 1619.38 3046.87 1146.17 1651.96 914.32 3624.45 2742.47 13
OpenMVS_ROBcopyleft52.78 1735.39 1727.98 1940.33 2146.81 1336.10 3924.79 1631.17 1638.08 22
CasMVSNet(base)35.32 1818.44 3346.56 1246.00 1852.64 813.32 3723.57 3041.04 17
IB-MVS56.42 1234.86 1928.22 1739.29 2543.59 2038.19 3223.59 1832.85 1236.09 27
GSE34.79 2023.63 2642.23 1946.31 1539.18 2923.20 2124.07 2841.20 16
PLCcopyleft56.13 1334.63 2128.70 1438.59 2836.78 3344.65 2126.03 1431.36 1434.33 30
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
AttMVS34.35 2228.45 1638.28 2940.03 2640.06 2623.67 1733.23 1134.75 29
LPCS33.63 2324.80 2239.52 2342.29 2239.01 3023.44 2026.15 2437.27 24
PVSNet_043.31 1933.35 2423.83 2439.69 2241.22 2338.75 3121.08 2526.59 2239.10 19
test_112633.11 2518.21 3543.04 1747.43 1145.33 1815.33 3321.10 3536.35 26
COLMAP_ROBcopyleft52.97 1633.06 2625.38 2138.18 3035.81 3545.04 1922.92 2327.84 2133.69 31
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
ACMH55.70 1432.91 2723.69 2539.06 2641.22 2336.76 3721.84 2425.54 2539.18 18
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
Pnet-new-32.65 2816.09 3843.68 1546.15 1739.98 2718.51 2813.67 4844.91 10
CIDER32.55 2925.75 2037.08 3339.65 2740.58 2520.91 2630.59 1831.00 35
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
PVSNet50.76 1832.44 3024.74 2337.57 3238.43 3036.66 3823.03 2226.46 2337.62 23
ANet30.02 3116.73 3738.88 2739.32 2844.06 2215.61 3217.85 3833.26 32
BP-MVSNet29.92 3220.53 2836.17 3437.89 3133.68 4317.55 2923.52 3136.95 25
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
P-MVSNet29.35 3320.98 2734.93 3534.55 3936.99 3520.06 2721.91 3433.25 33
ANet-0.7529.06 3413.44 4139.47 2439.32 2844.06 2210.78 4216.11 4235.02 28
test_120528.37 3519.29 3134.43 3634.50 4041.99 2415.81 3122.78 3226.79 40
Pnet_fast27.80 3612.25 4238.17 3137.73 3247.91 137.47 4817.03 3928.88 37
MVSNet27.45 3718.54 3233.39 3829.32 4247.13 1413.18 3823.89 2923.72 41
A-TVSNet + Gipumacopyleft26.68 3816.93 3633.18 3931.29 4139.82 2817.28 3016.57 4128.43 38
test_112423.88 398.92 5033.86 3735.80 3636.79 3610.99 416.84 5628.99 36
R-MVSNet23.84 4018.30 3427.54 4227.82 4333.79 4214.57 3522.03 3321.00 44
CasMVSNet(SR_B)23.80 4119.79 2926.47 4424.88 4437.51 3414.94 3424.65 2617.01 46
Pnet-blend21.89 429.21 4830.34 4034.58 3734.44 403.76 5914.67 4522.00 42
Pnet-blend++21.89 429.21 4830.34 4034.58 3734.44 403.76 5914.67 4522.00 42
MVSNet_plusplus19.10 446.64 5627.41 4336.63 3414.19 523.12 6110.15 5031.41 34
MVSCRF18.49 4512.01 4322.81 4520.51 4732.84 449.61 4314.41 4715.08 49
MVSNet_++16.58 4610.34 4620.75 4622.25 4511.60 551.64 6419.05 3628.39 39
Snet16.01 4711.97 4418.70 4722.09 4617.12 508.46 4615.48 4416.90 47
A1Net14.51 4814.40 4014.59 4813.30 4912.51 5311.95 3916.85 4017.96 45
CPR_FA14.43 4914.90 3914.12 5011.38 5014.82 5111.91 4017.89 3716.15 48
PMVScopyleft28.69 2110.60 504.70 6114.53 4910.82 5430.05 466.66 532.74 622.71 60
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
F/T MVSNet+Gipuma9.78 518.13 5110.88 569.23 5611.92 548.81 447.45 5411.49 50
firsttry9.57 527.40 5411.02 5511.05 5110.54 586.97 517.82 5311.49 50
example9.54 535.97 6011.92 5113.40 4820.72 476.89 525.05 601.63 65
MVSNet + Gipuma9.47 547.91 5210.52 578.91 5811.46 568.80 457.01 5511.18 53
hgnet9.31 556.52 5711.17 5310.91 5220.13 487.15 495.89 572.47 62
DPSNet9.31 556.52 5711.17 5310.91 5220.13 487.15 495.89 572.47 62
MVEpermissive17.77 229.03 579.66 478.62 586.03 6111.40 578.30 4711.02 498.42 54
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
Pnet-eth8.45 5810.49 457.10 619.98 552.92 665.14 5615.84 438.40 55
unMVSv16.85 596.06 597.37 607.56 599.33 596.35 545.76 595.22 57
CMPMVSbinary42.80 206.77 600.01 6711.27 521.93 6731.86 450.03 670.00 670.00 67
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
test_1120copyleft6.40 614.10 627.94 596.69 605.86 614.95 573.26 6111.26 52
metmvs_fine4.94 627.02 553.56 643.18 653.86 634.62 589.43 523.63 58
RMVSNet4.93 637.73 533.06 653.93 633.22 655.95 559.52 512.04 64
Cas-MVS_preliminary4.12 641.79 645.67 624.49 624.19 622.85 620.74 658.32 56
unMVSmet3.60 652.44 634.38 633.82 646.35 602.69 632.18 632.97 59
confMetMVS2.34 661.33 653.02 663.03 663.47 641.14 651.52 642.56 61
FADENet0.06 670.04 660.07 670.12 680.08 670.05 660.03 660.02 66
dnet0.00 680.00 680.00 680.00 690.00 680.00 680.00 670.00 67
UnsupFinetunedMVSNet9.23 56