Low-res many-view results

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.

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	all	low-res many-view	indoor	outdoor	delivery area	electro	forest	playground	terrains

ACMP	56.21 5	46.95 1	48.21 1	46.11 2	35.74 2	49.62 2	60.42 7	28.28 3	60.67 4
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
BP-MVSNet		45.71 2	45.99 4	45.52 4	36.41 1	51.51 1	63.02 3	22.04 11	55.56 12
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
TAPA-MVS(SR)		45.32 3	44.43 6	45.91 3	33.62 4	45.57 5	64.25 2	27.92 5	55.25 14

COLMAP(SR)		45.20 4	43.53 10	46.32 1	33.64 3	42.84 7	65.41 1	30.70 1	53.42 19

ACMH+	47.85 12	44.52 5	43.79 8	45.00 5	29.83 7	46.18 4	60.45 6	28.39 2	57.76 8

DeepC-MVS	60.65 2	43.29 6	46.59 3	41.10 10	30.52 5	40.64 10	56.08 11	26.58 10	62.65 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMM	53.73 9	42.86 7	42.29 11	43.24 8	25.37 12	43.20 6	58.55 8	27.97 4	59.20 7
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
DeepC-MVS_fast	60.18 3	42.85 8	44.64 5	41.66 9	29.91 6	40.32 11	57.24 10	27.41 7	59.38 6
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepPCF-MVS	62.48 1	42.76 9	47.85 2	39.37 11	28.07 9	38.72 12	52.59 18	26.81 9	67.63 1

IB-MVS	53.15 10	42.26 10	39.58 14	44.04 7	25.03 14	48.21 3	62.12 5	21.79 12	54.14 17
Christian Sormann, Mattia Rossi, Andreas Kuhn and Friedrich Fraundorfer: IB-MVS: An Iterative Algorithm for Deep Multi-View Stereo based on Binary Decisions. BMVC 2021
ACMH	47.82 13	41.87 11	38.07 16	44.41 6	21.70 20	42.80 8	62.67 4	27.77 6	54.44 16
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PCF-MVS	55.99 6	39.56 12	43.75 9	36.76 15	27.84 10	34.98 17	47.93 20	27.37 8	59.66 5
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
3Dnovator	58.39 4	38.67 13	40.68 12	37.34 13	23.82 19	37.03 14	55.50 13	19.48 15	57.54 9

TAPA-MVS	47.92 11	38.21 14	44.43 6	34.07 19	25.77 11	26.04 25	54.71 14	21.47 13	63.09 2
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
3Dnovator+	55.76 7	37.74 15	39.02 15	36.88 14	21.49 21	35.52 15	56.01 12	19.10 17	56.56 10

OpenMVS	55.62 8	37.55 16	40.35 13	35.68 16	25.18 13	35.46 16	53.95 15	17.65 22	55.52 13

CasMVSNet(SR_B)		37.08 17	35.07 21	38.42 12	24.93 15	41.34 9	57.42 9	16.52 23	45.21 24

CasMVSNet(base)		34.90 18	34.15 23	35.40 17	24.03 18	37.15 13	53.88 16	15.17 25	44.27 26

A-TVSNet + Gipuma		33.59 19	32.80 25	34.12 18	19.10 22	30.61 18	52.47 19	19.28 16	46.49 21

COLMAP(base)		33.41 20	37.01 19	31.02 21	18.89 23	29.06 19	43.92 22	20.07 14	55.12 15

GSE		32.30 21	37.29 17	28.97 23	28.18 8	26.11 24	42.34 23	18.45 20	46.40 22

LPCS		31.44 22	35.26 20	28.90 24	24.11 17	28.43 20	40.15 24	18.12 21	46.40 22

PLC	44.22 14	31.44 22	37.10 18	27.66 25	17.76 25	26.38 23	37.68 25	18.93 18	56.45 11
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
CIDER		29.50 24	24.02 28	33.15 20	13.56 26	28.04 21	52.69 17	18.72 19	34.49 30
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
unsupervisedMVS_cas		29.41 25	29.47 26	29.37 22	18.47 24	26.45 22	47.65 21	13.99 26	40.47 27

LTVRE_ROB	32.83 17	26.93 26	33.04 24	22.86 26	12.48 27	25.75 26	29.17 28	13.65 27	53.60 18
Andreas Kuhn, Heiko HirschmÃ¼ller, Daniel Scharstein, Helmut Mayer: A TV Prior for High-Quality Scalable Multi-View Stereo Reconstruction. International Journal of Computer Vision 2016
CasMVSNet(SR_A)		25.76 27	35.07 21	19.55 28	24.93 15	10.61 28	31.51 27	16.52 23	45.21 24

COLMAP_ROB	34.79 15	24.75 28	28.33 27	22.36 27	10.04 28	19.30 27	36.90 26	10.90 28	46.61 20
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
MVE	10.35 19	11.88 29	20.48 29	6.15 29	2.93 29	9.30 29	5.03 30	4.12 29	38.02 28
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
example		9.57 30	16.38 31	5.02 30	2.25 32	1.90 30	9.20 29	3.97 30	30.52 31

PMVS	18.18 18	8.40 31	17.69 30	2.21 33	0.86 33	1.25 33	4.14 33	1.25 31	34.52 29
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
hgnet		5.93 32	11.26 33	2.37 31	2.78 30	1.60 31	4.42 31	1.09 32	19.73 33

DPSNet		5.93 32	11.26 33	2.37 31	2.78 30	1.60 31	4.42 31	1.09 32	19.73 33

CMPMVS	33.64 16	5.69 34	14.19 32	0.03 34	0.00 34	0.00 34	0.09 34	0.00 34	28.38 32
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
dnet		0.00 35	0.00 35	0.00 35	0.00 34	0.00 34	0.00 35	0.00 34	0.00 35

Low-resolution many-view benchmark