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 : F1 score [%]
Tolerance : 20cm
Method
Info all low-res indoor outdoor lakeside sand box storage room storage room 2 tunnel
DeepPCF-MVS 96.37 2 92.61 1 89.11 1 94.94 1 95.58 1 95.07 1 85.62 1 92.60 1 94.17 1 DeepC-MVS_fast 96.44 1 91.83 2 87.74 3 94.57 2 95.34 3 94.64 2 84.48 4 91.00 4 93.72 3 DeepC-MVS 96.23 3 91.57 3 87.58 4 94.23 3 94.49 5 94.42 3 83.58 6 91.59 2 93.78 2 TAPA-MVS 93.98 7 90.68 4 88.05 2 92.44 7 92.42 10 93.52 8 85.24 2 90.85 6 91.36 14 Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo . ICCV 2019 TAPA-SR 90.57 5 87.48 5 92.63 6 92.64 8 93.17 11 83.91 5 91.04 3 92.10 12 PCF-MVS 93.45 10 90.50 6 85.08 13 94.11 4 95.54 2 93.44 10 82.27 11 87.88 15 93.34 4 Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction . GCPR 2019 CMAPSRf 90.17 7 85.79 11 93.08 5 92.63 9 93.87 6 82.51 10 89.07 13 92.74 6 PLC 95.07 4 89.96 8 87.00 6 91.94 12 90.65 17 92.68 15 83.01 8 90.98 5 92.48 7 Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency . Pacific Graphics 2019 CMAPbl 89.91 9 86.37 8 92.28 9 91.56 13 92.95 13 82.53 9 90.21 7 92.33 8 ACMH+ 92.99 13 89.84 10 85.96 10 92.43 8 92.28 11 92.84 14 81.89 12 90.03 8 92.17 10 GSE 89.80 11 86.71 7 91.87 13 90.87 16 92.62 16 83.49 7 89.92 9 92.11 11 ACMM 93.85 8 89.42 12 85.42 12 92.10 11 91.09 15 92.98 12 81.16 14 89.68 10 92.23 9 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 ACMP 93.49 9 89.04 13 84.30 16 92.21 10 90.34 18 93.47 9 79.54 18 89.07 13 92.81 5 Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo . AAAI 2020 LTVRE_ROB 92.95 14 89.00 14 86.32 9 90.78 19 91.67 12 89.54 22 84.87 3 87.77 16 91.13 16 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 92.88 15 88.89 15 84.55 14 91.78 14 91.37 14 92.41 17 79.88 16 89.21 11 91.56 13 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 COLMAP_ROB 93.27 11 88.32 16 84.41 15 90.93 18 89.38 19 92.11 18 79.72 17 89.10 12 91.28 15 Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo . ECCV 2016 3Dnovator 94.51 5 87.43 17 80.90 20 91.78 14 94.76 4 94.22 4 78.96 19 82.85 20 86.35 19 3Dnovator+ 94.38 6 87.07 18 80.26 21 91.61 16 94.34 6 94.17 5 78.01 20 82.52 21 86.34 20 A-TVSNet + Gipuma 86.96 19 82.91 18 89.65 20 88.10 20 91.71 19 80.23 15 85.60 19 89.15 17 OpenMVS 93.04 12 86.34 20 79.08 22 91.18 17 93.50 7 93.77 7 76.56 21 81.60 24 86.26 21 LPCS 85.52 21 83.61 17 86.79 21 84.99 23 88.49 24 81.34 13 85.87 18 86.89 18 R-MVSNet 81.26 22 77.44 23 83.81 24 82.82 24 86.25 26 74.52 23 80.36 25 82.37 22 CIDER 80.48 23 72.14 26 86.05 22 87.66 21 89.77 20 71.75 24 72.53 30 80.72 24 Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume . AAAI 2020 OpenMVS_ROB 86.42 16 80.25 24 71.88 28 85.82 23 87.52 22 89.56 21 70.53 25 73.24 29 80.39 25 ANet-0.75 77.73 25 74.78 25 79.70 27 77.01 30 84.71 29 67.34 29 82.21 23 77.39 28 P-MVSNet 77.69 26 81.61 19 75.07 33 71.91 35 77.33 34 76.51 22 86.71 17 75.98 30 CPR_FA 74.23 27 72.06 27 75.68 31 74.30 34 70.69 40 68.60 27 75.52 28 82.05 23 ANet 74.07 28 68.26 30 77.94 30 77.01 30 84.71 29 64.75 30 71.76 31 72.09 32 AttMVS 72.35 29 68.13 31 75.16 32 70.98 39 80.63 31 68.09 28 68.16 32 73.87 31 Pnet-new- 70.55 30 53.66 41 81.81 26 81.27 26 85.23 28 51.15 38 56.17 45 78.92 27 cscdsrtA 69.66 31 55.07 37 79.38 28 80.08 27 87.62 25 46.56 43 63.58 38 70.44 33 Pnet_fast 68.89 32 47.01 49 83.47 25 81.84 25 89.30 23 38.61 51 55.42 46 79.28 26 cscd 68.63 33 53.70 40 78.58 29 79.92 28 85.96 27 44.83 45 62.57 40 69.87 34 A1Net 68.49 34 75.64 24 63.72 40 63.53 45 51.39 50 68.88 26 82.41 22 76.23 29 Pnet-blend++ 64.02 35 52.13 45 71.95 34 75.81 32 78.00 32 40.66 48 63.61 36 62.05 38 Pnet-blend 64.02 35 52.13 45 71.95 34 75.81 32 78.00 32 40.66 48 63.61 36 62.05 38 unMVSv1 63.91 37 62.59 33 64.80 39 66.39 42 68.00 43 61.62 31 63.55 39 60.01 41 MVSNet 63.58 38 56.25 35 68.47 38 66.00 43 71.12 37 48.36 40 64.13 33 68.29 36 MVSCRF 63.13 39 52.16 44 70.45 36 70.18 40 75.83 36 51.17 37 53.16 49 65.33 37 RMVSNet 62.69 40 69.01 29 58.48 48 69.15 41 58.56 46 59.72 32 78.30 26 47.75 49 Snet 60.77 41 47.18 48 69.82 37 77.90 29 61.73 45 39.61 50 54.76 47 69.82 35 Pnet-eth 60.43 42 65.87 32 56.80 50 63.59 44 45.39 51 55.58 34 76.16 27 61.43 40 hgnet 59.89 43 54.16 38 63.70 41 71.39 36 70.87 38 52.09 35 56.24 43 48.85 46 DPSNet 59.89 43 54.16 38 63.70 41 71.39 36 70.87 38 52.09 35 56.24 43 48.85 46 MVE 62.14 18 59.41 45 60.38 34 58.77 47 50.39 52 69.00 42 56.98 33 63.78 35 56.91 45 Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment . EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014) MVSNet + Gipuma 57.07 46 53.08 42 59.73 46 62.91 48 57.42 48 48.41 39 57.74 41 58.88 42 cscdsrtB 56.87 47 55.95 36 57.48 49 56.06 50 67.85 44 47.86 42 64.05 34 48.53 48 F/T MVSNet+Gipuma 55.78 48 49.83 47 59.76 45 63.10 46 57.51 47 46.34 44 53.31 48 58.66 43 example 55.06 49 44.75 50 61.93 43 71.21 38 69.85 41 42.82 47 46.68 50 44.73 50 firsttry 54.62 50 52.74 43 55.88 51 54.90 51 55.69 49 48.02 41 57.46 42 57.06 44 PMVS 61.03 19 49.24 51 32.69 51 60.27 44 62.70 49 76.26 35 42.99 46 22.38 51 41.86 51 Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis . PAMI (2010) CMPMVS 66.06 17 9.32 52 0.44 53 15.23 52 6.17 53 39.51 52 0.89 53 0.00 53 0.00 53 M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces . CVPR 2011 FADENet 4.02 53 4.30 52 3.83 53 6.03 54 3.71 53 5.02 52 3.58 52 1.74 52 dnet 0.00 54 0.00 54 0.00 54 0.00 55 0.00 54 0.00 54 0.00 53 0.00 53 UnsupFinetunedMVSNet 63.10 46
