This table lists the benchmark results for the low-res many-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 : Completeness [%]
Tolerance : 2cm
Method
Info all low-res indoor outdoor lakeside sand box storage room storage room 2 tunnel
A1Net 62.59 1 54.91 3 67.70 3 68.95 1 72.10 4 52.15 2 57.67 3 62.06 7 ACMP 71.68 7 62.33 2 51.03 7 69.86 1 66.41 2 74.93 1 47.36 7 54.70 7 68.24 1 Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo . AAAI 2020 A-TVSNet + Gipuma 61.02 3 56.34 1 64.14 7 64.32 4 66.91 15 52.09 3 60.60 1 61.19 10 DeepPCF-MVS 81.17 1 60.64 4 53.51 5 65.40 5 64.27 5 69.40 10 51.28 5 55.74 6 62.53 5 ACMH+ 65.35 12 59.83 5 46.18 12 68.93 2 65.70 3 74.24 2 46.34 11 46.01 12 66.83 2 TAPA-MVS 70.22 9 58.89 6 51.21 6 64.01 8 62.59 7 66.45 17 46.31 12 56.11 5 62.99 4 Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo . ICCV 2019 DeepC-MVS 77.92 3 58.46 7 50.82 8 63.55 10 62.52 8 66.82 16 47.58 6 54.06 8 61.30 9 PCF-MVS 73.15 6 58.42 8 54.11 4 61.29 13 61.25 11 64.94 18 51.80 4 56.43 4 57.66 11 Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction . GCPR 2019 ACMM 69.62 10 58.27 9 46.31 11 66.24 4 60.67 13 72.46 3 42.63 13 50.00 10 65.60 3 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 DeepC-MVS_fast 78.06 2 58.14 10 48.66 9 64.46 6 63.44 6 67.77 13 46.93 9 50.38 9 62.16 6 PLC 68.80 11 53.38 11 47.19 10 57.51 19 57.17 18 61.22 19 46.82 10 47.56 11 54.14 15 Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency . Pacific Graphics 2019 3Dnovator+ 73.60 5 53.20 12 38.24 14 63.18 11 62.18 9 70.72 5 36.81 14 39.68 16 56.64 13 3Dnovator 73.91 4 52.81 13 38.17 15 62.58 12 60.95 12 69.63 9 35.98 17 40.36 15 57.16 12 ANet-0.75 52.71 14 55.63 2 50.76 21 44.38 22 59.75 20 53.05 1 58.21 2 48.17 19 ACMH 63.93 13 52.68 15 35.79 19 63.93 9 60.12 16 70.07 8 34.99 18 36.59 19 61.61 8 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 OpenMVS 70.45 8 51.01 16 37.34 16 60.13 14 56.52 19 67.91 12 33.72 19 40.96 14 55.96 14 CIDER 49.33 17 33.81 20 59.69 16 61.47 10 68.26 11 31.50 20 36.11 20 49.34 18 Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume . AAAI 2020 OpenMVS_ROB 61.12 14 48.48 18 31.63 21 59.71 15 59.03 17 67.15 14 31.08 21 32.18 22 52.95 16 COLMAP_ROB 57.96 16 45.89 19 37.03 18 51.79 20 49.43 20 56.42 25 36.78 15 37.28 18 49.53 17 Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo . ECCV 2016 LTVRE_ROB 59.60 15 43.60 20 37.31 17 47.80 22 45.94 21 49.52 30 36.19 16 38.43 17 47.93 20 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 unMVSv1 43.21 21 44.62 13 42.27 26 40.66 25 51.71 28 47.08 8 42.16 13 34.44 27 Pnet_fast 41.99 22 16.05 31 59.29 17 60.27 14 70.55 6 9.62 37 22.48 30 47.04 21 Snet 41.99 22 16.05 31 59.29 17 60.27 14 70.55 6 9.62 37 22.48 30 47.04 21 P-MVSNet 38.28 24 29.70 23 44.00 25 39.20 26 50.96 29 29.12 22 30.29 24 41.85 23 R-MVSNet 37.16 25 26.49 24 44.27 24 42.37 24 52.02 27 24.06 24 28.93 26 38.41 26 AttMVS 37.07 26 30.02 22 41.77 27 34.23 27 52.58 26 28.11 23 31.93 23 38.51 25 ANet 37.00 27 21.13 27 47.57 23 44.38 22 59.75 20 18.53 26 23.72 28 38.59 24 firsttry 32.39 28 25.47 25 36.99 28 32.51 28 46.62 31 21.58 25 29.36 25 31.86 29 MVSNet 27.67 29 18.22 29 33.96 29 26.28 31 43.46 33 13.67 31 22.78 29 32.14 28 MVE 24.84 19 26.04 30 20.59 28 29.66 32 22.29 34 42.00 34 15.49 30 25.70 27 24.70 31 Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment . EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014) MVSCRF 24.97 31 13.50 33 32.61 30 26.31 30 43.85 32 12.02 33 14.98 33 27.67 30 example 24.50 32 16.56 30 29.80 31 27.47 29 56.53 24 17.69 27 15.43 32 5.40 37 hgnet 22.90 33 13.03 34 29.49 33 25.13 32 56.72 22 16.70 28 9.35 36 6.62 35 DPSNet 22.90 33 13.03 34 29.49 33 25.13 32 56.72 22 16.70 28 9.35 36 6.62 35 F/T MVSNet+Gipuma 19.68 35 11.34 36 25.23 35 16.34 35 38.41 35 11.15 35 11.54 34 20.95 32 MVSNet + Gipuma 19.59 36 11.32 37 25.11 36 16.29 37 38.21 36 11.18 34 11.45 35 20.83 33 RMVSNet 15.61 37 22.67 26 10.90 38 13.04 39 14.34 39 12.27 32 33.07 21 5.30 38 PMVS 26.43 18 13.46 38 7.26 38 17.59 37 14.19 38 30.85 37 10.42 36 4.11 38 7.73 34 Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis . PAMI (2010) CMPMVS 48.56 17 4.84 39 0.02 39 8.06 39 1.22 40 22.96 38 0.03 39 0.00 39 0.00 39 M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces . CVPR 2011 dnet 0.00 40 0.00 40 0.00 40 0.00 41 0.00 40 0.00 40 0.00 39 0.00 39 UnsupFinetunedMVSNet 16.34 35
