This table lists the benchmark results for the high-res multi-view scenario.
The following metrics are evaluated:
Accuracy [%]: The fraction of the reconstruction which is closer to the ground truth than the evaluation threshold distance (*). Larger is better.Completeness [%]: The fraction of the ground truth which is closer to the reconstruction than the evaluation threshold distance (*). Larger is better.F1 score [%]: The harmonic mean of accuracy and completeness, used to rank methods based on both metrics. Larger is better.Time [s]: The runtime of the method. Smaller is better.
(*) 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 .
training datasets , ground truth and accuracy / completeness visualizations are also available. The visualizations may not work with mobile browsers.
Set : Test
Metric : Time [s]
Method
Info all high-res indoor outdoor botani. boulde. bridge door exhibi. lectur. living. lounge observ. old co. statue terrac.
patch_net 492.52 1 551.22 1 316.40 2 400.20 3 357.90 2 1277.10 1 115.80 3 1034.60 2 315.90 2 874.00 2 155.10 6 402.20 3 629.50 2 158.80 2 189.10 2 ACMH+ 729.97 1 661.38 2 754.00 2 383.51 3 368.96 2 376.38 3 2483.74 4 82.42 2 1089.25 3 333.39 3 1219.68 4 66.53 2 573.60 5 957.86 6 184.14 3 200.56 3 Gipuma 689.75 3 767.00 3 458.00 4 613.00 9 549.00 6 1943.00 2 187.00 10 1230.00 4 518.00 6 1092.00 3 203.00 12 564.00 4 892.00 5 225.00 4 261.00 4 S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion . ICCV 2015 PVSNet 1207.94 4 829.56 4 937.18 4 506.69 6 645.55 10 585.72 7 2477.98 3 130.52 4 1578.29 5 509.50 5 1460.01 5 194.90 9 637.19 6 1170.78 7 267.07 7 297.17 6 PMVS 863.65 2 957.08 5 1054.11 5 666.00 10 401.00 4 495.00 4 3790.00 7 279.00 13 1607.00 7 521.00 7 1745.00 7 90.00 3 1105.00 12 639.00 3 415.00 12 398.00 11 Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis . PAMI (2010) ACMH 993.80 3 967.92 6 1106.11 7 553.33 7 494.00 5 534.00 5 3581.00 6 136.00 5 1585.00 6 478.00 4 1871.00 8 90.00 3 839.00 7 1460.00 9 260.00 6 287.00 5 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 PVSNet_0 1661.03 7 968.54 7 1066.82 6 673.70 11 706.94 12 695.69 9 2669.02 5 153.52 7 2108.17 11 605.54 9 1679.14 6 213.26 14 886.75 8 1236.76 8 229.01 5 438.67 12 ACMP 1246.57 5 1086.24 8 1235.17 8 639.45 8 596.53 7 639.91 8 3904.24 8 152.04 6 1851.54 8 566.58 8 1997.95 9 138.50 5 938.30 9 1605.46 11 303.70 9 340.14 8 Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo . AAAI 2020 ACMM 1311.08 6 1165.06 9 1332.72 9 662.07 9 584.90 6 715.99 10 4373.59 10 153.99 8 1880.27 9 709.02 11 2316.97 10 188.63 8 939.89 10 1484.30 10 302.78 8 330.34 7 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 MG-MVS 1256.50 10 1432.00 10 730.00 12 610.00 8 760.00 11 4720.00 11 169.00 9 1920.00 10 723.00 12 2510.00 13 201.00 11 1050.00 11 1645.00 12 390.00 11 380.00 9 FPMVS 1407.00 11 1495.59 11 1141.23 14 648.01 11 893.85 13 4124.00 9 503.81 23 3219.04 15 742.72 13 2498.84 12 157.98 7 1727.83 16 832.93 4 732.98 17 802.00 16 DeepC-MVS_fast 2190.19 10 1535.59 12 1886.82 13 481.89 5 847.73 13 830.16 12 7943.84 19 189.23 11 2398.83 12 691.73 10 2341.13 11 209.16 13 230.25 2 1999.25 13 360.46 10 385.25 10 COLMAP_ROB 2144.59 9 1658.33 13 1869.33 12 1025.33 13 1030.00 14 1023.00 14 5311.00 13 359.00 14 3044.00 14 937.00 14 2892.00 14 332.00 17 1440.00 13 2312.00 14 607.00 13 613.00 14 Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo . ECCV 2016 F-COLMAP 1878.17 14 2113.11 14 1173.33 16 1231.00 15 1174.00 15 6297.00 14 392.00 16 3356.00 17 1087.00 15 3082.00 15 361.00 19 1670.00 15 2523.00 16 689.00 14 676.00 15 OpenMVS_ROB 1936.51 8 1959.51 15 2224.88 16 1163.39 15 1253.91 16 1434.22 18 7065.71 15 272.76 12 2472.29 13 1647.54 25 3538.72 17 200.37 10 1496.48 14 2575.17 17 997.45 25 559.48 13 CMPMVS 2623.62 13 1983.08 16 2156.14 15 1463.89 21 1414.80 18 1517.00 19 5239.20 12 867.40 31 3310.30 16 1128.70 17 3489.90 16 366.75 20 2030.20 21 2427.60 15 1160.60 27 844.46 20 M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces . CVPR 2011 PLC 2858.82 14 2192.92 17 2474.00 17 1349.67 18 1407.00 17 1356.00 17 7164.00 16 475.00 21 3970.00 19 1237.00 20 3765.00 20 451.00 21 1891.00 18 3006.00 19 791.00 18 802.00 16 Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency . Pacific Graphics 2019 OpenMVS 2352.88 11 2263.08 18 2530.89 18 1459.67 20 1767.00 22 1591.00 22 7932.00 18 399.00 17 3643.00 18 1176.00 18 3569.00 18 297.00 16 1927.00 20 3302.00 20 693.00 15 861.00 21 PCF-MVS 4335.85 18 2272.23 19 2597.03 19 1297.84 17 1701.20 19 1332.37 16 8105.11 20 451.45 18 4031.84 20 1099.10 16 3966.50 21 467.45 22 1734.05 17 2701.68 18 848.94 21 827.11 19 Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction . GCPR 2019 3Dnovator 2482.06 12 2531.17 20 2839.56 20 1606.00 22 1732.00 20 1557.00 21 8375.00 22 483.00 22 4370.00 21 1371.00 22 4200.00 22 350.00 18 2207.00 22 3877.00 24 798.00 19 1054.00 26 AdaColmap 2570.77 21 2948.61 21 1437.24 19 1873.64 23 1526.60 20 7819.77 17 373.97 15 4970.45 24 1479.29 23 4648.46 24 525.80 25 1904.97 19 4150.42 25 695.69 16 880.14 22 DeepPCF-MVS 3452.42 16 2670.71 22 3021.57 22 1618.12 23 2377.67 26 1751.39 24 8915.70 24 567.44 24 4569.74 22 1236.84 19 4615.63 23 513.23 24 2291.32 23 3515.48 21 882.39 23 811.66 18 LF4IMVS 2699.50 23 3049.67 23 1649.00 24 1755.00 21 1689.00 23 8330.00 21 464.00 20 4997.00 25 1528.00 24 4763.00 25 568.00 26 2296.00 24 4215.00 26 827.00 20 962.00 24 3Dnovator+ 3414.41 15 3146.67 24 3537.00 24 1975.67 26 2310.00 25 2005.00 27 10100.00 25 626.00 26 5753.00 26 1803.00 27 5179.00 26 487.00 23 2622.00 26 4550.00 27 1025.00 26 1300.00 28 DeepC-MVS 3917.13 17 3155.06 25 3601.83 25 1814.72 25 2515.52 27 1941.53 25 10432.50 26 589.42 25 4789.45 23 1298.65 21 7677.31 29 578.35 27 2589.53 25 3611.78 23 923.52 24 913.10 23 TAPA-MVS 6309.84 19 3374.90 26 3806.99 26 2078.63 28 2535.47 28 2499.25 28 8816.16 23 456.05 19 6255.60 28 1664.52 26 6704.88 28 629.95 28 2768.38 27 6343.93 30 856.35 22 968.25 25 Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo . ICCV 2019 MAR-MVS 3973.16 27 4387.92 27 2728.86 29 3345.27 29 3341.39 30 11403.90 27 798.76 29 6555.62 29 2569.29 29 6514.86 27 834.97 32 3237.43 29 5891.08 29 1577.54 30 1607.76 29 Zhenyu Xu, Yiguang Liu, Xuelei Shi, Ying Wang, Yunan Zheng: MARMVS: Matching Ambiguity Reduced Multiple View Stereo for Efficient Large Scale Scene Reconstruction . CVPR 2020 AP-MVS 5113.33 28 5849.36 28 2905.24 30 5005.38 31 3177.54 29 15238.40 28 828.78 30 9372.12 30 3019.68 30 8934.18 30 1103.28 33 3848.64 30 7738.80 32 1403.64 29 1689.54 31 CLD-MVS 7338.25 29 8341.00 29 4330.00 31 4902.00 30 4991.00 31 23318.00 29 728.00 28 14437.00 33 4378.00 32 13379.00 31 824.00 31 5708.00 31 11429.00 33 1674.00 31 2291.00 32 Zhaoxin Li, Wangmeng Zuo, Zhaoqi Wang, Lei Zhang: Confidence-based Large-scale Dense Multi-view Stereo . IEEE Transaction on Image Processing, 2020 LS3D 7684.71 30 8776.23 30 4410.16 32 5298.55 32 5631.53 32 34329.90 31 1504.57 32 11554.00 32 3896.41 31 15111.00 32 812.19 30 5985.63 32 3516.20 22 2963.23 32 1613.31 30 MVE 10021.00 20 10550.67 31 10002.44 31 12195.33 33 6242.00 33 8768.00 33 30896.00 30 4451.00 33 10331.00 31 6247.00 33 18296.00 33 283.00 15 21484.00 33 7364.00 31 5912.00 33 6334.00 33 Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment . EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014) LTVRE_ROB 20565.47 21 23378.08 32 22805.00 32 25097.33 34 15500.00 34 19201.00 34 42712.00 32 6055.00 34 38388.00 34 16926.00 34 38534.00 34 7071.00 34 35653.00 34 29475.00 34 10584.00 34 20438.00 35 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 MSDG 32606.17 33 34443.44 33 27094.33 35 29249.00 35 25246.00 35 51144.00 33 6705.00 35 42886.00 35 22888.00 35 48534.00 35 13263.00 35 37507.00 35 79298.00 35 16024.00 35 18530.00 34 PNet_i23d 10000000.00 34 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 wuykxyi23d 10000000.00 34 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 ANet_high 10000000.00 34 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 wuyk23d 10000000.00 34 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 34 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 10000000.00 36 DeepMVS_CX 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 1.00 1 ITE_SJBPF 2014.00 27 1998.00 24 1992.00 26 701.00 27 5989.00 27 1812.00 28 3612.00 19 650.00 29 2852.00 28 4601.00 28 1195.00 28 1198.00 27
