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 : F1 score [%]
Tolerance : 5cm
Method
Info all high-res indoor outdoor botani. boulde. bridge door exhibi. lectur. living. lounge observ. old co. statue terrac.
DeepPCF-MVS 89.96 1 94.20 1 94.77 1 92.49 10 96.52 8 80.00 21 94.00 18 97.08 1 90.05 2 95.65 1 97.29 1 89.66 1 98.97 8 93.95 1 98.71 1 98.50 2 DeepC-MVS_fast 89.26 2 93.75 2 93.41 2 94.79 2 97.70 1 86.68 2 95.98 4 96.92 3 89.90 3 91.05 3 96.94 2 83.92 3 99.32 2 89.75 3 98.50 2 98.36 3 DeepC-MVS 88.90 3 93.43 3 93.15 3 94.26 4 96.75 5 86.08 4 95.41 6 96.80 4 88.89 6 91.34 2 96.86 3 84.30 2 99.28 3 90.42 2 97.61 5 97.42 6 3Dnovator+ 87.14 4 92.42 4 91.37 5 95.55 1 95.63 12 88.73 1 97.07 2 96.77 5 90.84 1 84.02 13 96.62 5 75.95 10 99.34 1 87.77 4 97.68 4 98.59 1 MG-MVS 91.77 5 91.70 4 92.00 13 97.08 4 80.03 20 93.60 19 95.18 17 87.85 8 90.89 4 96.47 6 82.06 4 98.36 13 85.07 10 97.04 6 97.62 5 3Dnovator 86.66 5 91.73 6 90.82 6 94.44 3 94.59 17 86.37 3 97.18 1 97.02 2 89.20 4 84.31 12 96.66 4 73.74 13 99.17 5 86.74 7 97.96 3 97.79 4 AP-MVS 91.12 7 90.53 7 92.88 7 95.68 11 83.67 8 94.57 12 95.91 11 88.94 5 89.39 5 96.14 7 76.02 9 98.23 14 85.77 8 92.40 14 96.75 7 MAR-MVS 90.30 8 89.37 8 93.07 6 96.61 7 84.48 6 95.68 5 95.67 14 82.36 14 87.85 6 92.85 19 76.63 8 98.80 10 80.01 12 96.68 7 95.91 9 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 AdaColmap 89.89 9 89.07 9 92.37 11 97.41 2 83.03 9 94.42 14 95.92 10 82.81 13 86.34 7 94.65 11 73.89 12 99.02 6 80.69 11 95.51 9 95.05 11 CLD-MVS 89.47 10 88.90 10 91.18 16 94.22 20 82.07 12 92.13 22 96.09 9 87.90 7 85.37 10 92.45 23 74.38 11 97.56 21 87.15 5 90.43 16 93.93 17 Zhaoxin Li, Wangmeng Zuo, Zhaoqi Wang, Lei Zhang: Confidence-based Large-scale Dense Multi-view Stereo . IEEE Transaction on Image Processing, 2020 ACMM 84.12 9 89.14 11 88.48 11 91.12 17 94.65 16 81.22 15 95.31 7 96.12 8 85.31 11 85.92 9 94.34 13 70.19 16 98.06 16 85.65 9 88.86 17 94.08 16 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 PLC 84.53 7 89.06 12 88.03 13 92.15 12 97.27 3 82.69 11 94.29 16 95.44 16 79.71 17 84.01 14 94.18 16 76.68 7 98.75 11 77.28 14 93.41 13 95.02 12 Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency . Pacific Graphics 2019 ACMP 84.23 8 89.01 13 88.35 12 90.99 18 94.73 14 81.27 14 95.07 9 95.89 12 86.48 9 83.67 15 94.30 15 69.33 19 97.99 17 87.10 6 88.55 18 93.72 18 Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo . AAAI 2020 OpenMVS 83.78 11 88.74 14 87.29 14 93.08 5 92.70 23 85.39 5 96.57 3 96.43 6 78.74 20 80.85 18 96.07 8 69.64 18 99.01 7 78.01 13 96.65 8 94.83 13 TAPA-MVS 84.62 6 88.16 15 87.01 15 91.62 14 96.64 6 80.65 17 94.39 15 96.21 7 76.38 21 86.19 8 95.44 10 79.75 5 98.08 15 62.75 27 95.29 10 96.13 8 Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo . ICCV 2019 F-COLMAP 87.95 16 86.80 16 91.40 15 96.35 9 80.88 16 94.73 10 95.45 15 79.65 18 82.04 17 94.61 12 71.13 15 98.50 12 76.24 16 91.05 15 94.80 15 LS3D 87.89 17 86.32 17 92.59 9 96.07 10 82.92 10 95.23 8 94.92 18 75.66 23 82.89 16 95.98 9 72.48 14 99.21 4 68.43 25 95.23 11 95.64 10 PCF-MVS 84.11 10 87.74 18 86.08 18 92.70 8 94.02 21 84.43 7 89.27 25 95.87 13 73.62 25 84.43 11 94.33 14 78.48 6 98.86 9 70.27 22 94.45 12 94.81 14 Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction . GCPR 2019 LTVRE_ROB 82.13 12 86.26 19 84.90 19 90.34 20 94.44 19 81.50 13 92.31 21 94.89 19 83.03 12 79.63 20 92.67 21 69.69 17 97.79 19 71.20 21 86.26 21 91.72 22 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 80.38 15 85.36 20 83.68 20 90.39 19 94.45 18 80.63 18 94.73 10 94.85 20 82.09 16 77.24 24 92.65 22 60.01 22 97.58 20 72.25 20 84.87 23 92.96 20 Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo . CVPR 2019 ACMH+ 81.04 13 85.05 21 83.46 21 89.82 22 94.66 15 79.37 23 94.44 13 94.12 21 82.19 15 78.04 22 92.82 20 58.23 24 97.54 22 73.77 18 82.90 26 92.54 21 MSDG 84.86 22 83.09 22 90.14 21 93.80 22 80.05 19 89.18 26 93.09 23 78.89 19 78.19 21 91.91 24 65.86 20 97.27 23 68.47 24 88.45 19 93.11 19 COLMAP_ROB 80.39 14 83.96 23 82.04 23 89.74 23 95.28 13 79.75 22 94.25 17 92.28 25 75.17 24 78.02 23 93.77 17 58.60 23 97.84 18 65.06 26 85.92 22 91.63 23 Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo . ECCV 2016 CMPMVS 59.16 18 80.52 24 79.20 24 84.48 26 83.98 29 67.63 28 89.95 23 93.84 22 64.79 29 66.81 28 91.14 25 57.93 25 95.17 26 76.25 15 88.10 20 90.65 25 M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces . CVPR 2011 LF4IMVS 80.37 25 79.07 25 84.27 28 86.64 28 69.87 27 89.39 24 91.05 26 76.38 21 74.97 25 90.00 26 47.85 28 94.25 27 74.55 17 80.82 27 88.69 26 OpenMVS_ROB 74.94 16 79.51 26 77.03 26 86.93 25 87.00 27 76.23 25 92.33 20 90.74 27 68.93 28 74.52 26 88.23 27 49.58 27 96.62 24 57.64 28 84.29 25 87.94 28 PVSNet_0 73.20 17 77.22 27 74.83 27 84.37 27 90.70 25 71.10 26 83.09 28 89.67 29 72.81 27 73.93 27 83.13 28 60.79 21 93.70 28 68.54 23 50.84 32 88.30 27 FPMVS 64.63 28 62.55 28 70.88 30 70.80 31 56.71 30 84.42 27 84.42 31 51.78 30 49.57 29 81.61 29 23.49 30 81.48 31 40.61 32 76.25 28 74.46 31 ANet_high 58.88 29 54.22 30 72.86 29 56.50 36 56.67 31 80.75 29 86.00 30 73.09 26 37.39 32 64.63 32 22.17 31 79.49 33 43.51 30 23.96 36 82.43 29 Gipuma 57.99 30 54.91 29 67.24 32 88.51 26 65.59 29 52.21 34 90.33 28 43.58 33 42.84 31 51.18 35 20.29 32 85.07 29 34.77 34 70.45 29 51.05 36 S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion . ICCV 2015 PMVS 47.18 19 52.22 31 48.46 31 63.48 33 45.72 37 46.20 35 73.41 30 78.31 33 41.03 34 30.06 34 65.68 31 6.05 34 83.43 30 30.04 35 65.86 30 60.80 33 Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis . PAMI (2010) wuykxyi23d 50.55 32 44.13 33 69.81 31 56.77 34 54.58 32 73.22 31 80.78 32 39.79 35 22.08 36 46.69 36 4.03 36 79.71 32 47.65 29 26.13 35 75.14 30 PNet_i23d 50.48 33 47.18 32 60.36 34 68.59 32 44.56 36 72.75 32 72.61 34 43.92 32 33.91 33 60.19 33 6.16 33 73.52 34 38.50 33 28.04 34 63.01 32 MVE 39.65 20 43.39 34 38.59 34 57.77 35 56.52 35 48.77 34 55.38 33 58.64 35 29.33 36 28.96 35 52.65 34 4.68 35 64.62 35 28.11 36 33.07 33 59.93 34 Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment . EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014) wuyk23d 21.27 35 20.48 35 23.63 37 68.59 32 36.41 37 49.57 35 6.85 37 9.37 37 7.89 37 4.46 37 4.03 36 31.37 37 17.47 37 16.07 37 3.12 37 DeepMVS_CX 56.31 36 74.23 30 51.81 33 56.67 36 44.85 31 48.54 30 75.16 30 27.87 29 58.74 36 40.92 31 52.22 31 58.39 35 ITE_SJBPF 88.24 24 91.88 24 77.05 24 92.92 24 85.54 10 80.13 19 93.30 18 57.29 26 96.20 25 72.46 19 84.71 24 91.49 24
