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 bysorted bysort bysort bysort bysort bysort bysort by
DeepPCF-MVS80.84 163.41 153.95 169.71 169.72 371.26 148.73 159.18 168.16 1
TAPA-MVS73.13 958.67 652.34 262.89 963.68 863.46 945.85 458.83 261.53 8
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
DeepC-MVS79.50 361.99 352.12 368.57 367.79 570.58 345.96 358.27 367.34 3
DeepC-MVS_fast79.62 262.35 251.75 469.41 269.38 470.78 246.85 256.66 468.08 2
3Dnovator76.31 559.36 548.19 566.81 570.04 267.59 544.27 552.10 762.80 7
PCF-MVS73.52 757.06 748.10 663.03 866.85 662.36 1043.32 852.89 559.86 10
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
3Dnovator+77.84 459.53 447.18 767.76 470.63 167.68 443.34 751.01 964.98 5
PLCcopyleft70.83 1153.50 1245.79 858.64 1356.31 1560.76 1241.70 949.88 1158.86 13
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
OpenMVScopyleft72.83 1056.18 945.66 963.19 764.67 765.13 839.60 1151.71 859.76 11
LTVRE_ROB69.57 1253.52 1145.46 1058.89 1158.76 1360.60 1343.91 647.01 1357.32 14
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
ACMP74.13 656.41 844.60 1164.28 661.46 1065.81 737.07 1352.13 665.56 4
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
ACMH+68.96 1352.02 1443.83 1257.49 1459.85 1151.27 2040.04 1047.62 1261.35 9
ACMM73.20 855.01 1043.19 1362.89 959.60 1266.07 635.89 1550.48 1063.01 6
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
COLMAP_ROBcopyleft66.92 1552.32 1342.45 1458.89 1156.18 1661.09 1138.61 1246.28 1459.41 12
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
AttMVS45.85 1839.26 1550.25 2049.36 2051.75 1934.83 1643.70 1549.63 17
OpenMVS_ROBcopyleft64.09 1648.56 1538.68 1655.15 1561.90 949.65 2236.02 1441.33 1853.89 16
CIDER47.21 1738.36 1753.11 1755.37 1755.10 1833.25 1943.46 1648.87 19
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
ACMH67.68 1447.97 1638.24 1854.45 1656.88 1450.08 2134.49 1741.99 1756.40 15
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
P-MVSNet44.46 1937.09 1949.37 2249.27 2149.30 2334.35 1839.83 1949.54 18
R-MVSNet36.87 2429.78 2041.61 2442.00 2346.99 2424.73 2134.83 2035.83 24
A-TVSNet + Gipuma42.12 2029.09 2150.80 1848.28 2256.53 1529.14 2029.04 2247.59 20
MVSNet38.33 2326.99 2245.89 2340.49 2457.57 1420.52 2333.47 2139.63 23
ANet39.53 2124.60 2349.48 2149.42 1955.75 1622.08 2227.13 2443.25 22
A1Net23.74 2823.95 2423.61 2921.43 3219.62 3419.93 2427.96 2329.78 25
ANet-0.7539.47 2222.61 2550.71 1949.43 1855.75 1618.21 2527.01 2546.96 21
MVSCRF28.32 2518.18 2635.09 2532.16 2744.37 2614.66 2721.69 2828.73 26
Pnet_fast23.85 2617.41 2728.15 2632.72 2524.90 3112.22 3422.60 2626.83 27
Snet23.85 2617.41 2728.15 2632.72 2524.90 3112.22 3422.60 2626.83 27
MVEpermissive26.22 1916.26 3416.97 2915.79 3611.75 3819.40 3514.45 2819.48 2916.21 32
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
RMVSNet11.09 3815.49 308.17 399.16 3910.05 3912.57 3318.40 305.29 37
F/T MVSNet+Gipuma17.31 3014.39 3119.26 3116.83 3419.67 3314.03 2914.75 3121.27 29
MVSNet + Gipuma16.91 3114.12 3218.77 3416.39 3619.13 3613.98 3014.27 3220.78 30
firsttry15.88 3512.57 3318.08 3517.80 3317.14 3711.54 3713.61 3319.31 31
hgnet16.45 3212.31 3419.21 3221.51 3030.31 2912.80 3111.82 355.81 35
DPSNet16.45 3212.31 3419.21 3221.51 3030.31 2912.80 3111.82 355.81 35
unMVSv113.05 3712.03 3613.73 3714.18 3716.18 3812.13 3611.93 3410.84 34
PMVScopyleft37.38 1821.09 2911.49 3727.48 2824.09 2844.44 2515.98 267.01 3813.92 33
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
example15.69 3610.14 3819.39 3023.73 2930.32 2811.10 389.17 374.13 38
CMPMVSbinary51.72 177.38 390.03 3912.27 382.37 4034.46 270.06 390.00 390.00 39
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
dnet0.00 400.00 400.00 400.00 410.00 400.00 400.00 390.00 39
UnsupFinetunedMVSNet16.83 34