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

DeepPCF-MVS	79.04 2	62.76 1	66.22 1	60.45 2	53.69 1	65.50 1	68.81 5	47.05 3	78.76 1

DeepC-MVS_fast	78.07 3	62.18 2	63.61 2	61.23 1	52.80 2	65.06 2	70.53 2	48.10 1	74.42 5

DeepC-MVS	79.08 1	61.47 3	63.07 3	60.40 3	49.90 3	64.01 3	69.80 4	47.39 2	76.24 2

3Dnovator+	75.73 4	58.40 4	59.88 5	57.41 4	46.92 7	62.89 4	70.54 1	38.80 13	72.84 6

3Dnovator	73.76 5	58.20 5	60.10 4	56.93 5	45.58 9	62.04 5	70.17 3	38.57 14	74.61 4

PCF-MVS	73.28 6	57.32 6	59.66 6	55.76 6	48.61 4	56.36 12	67.24 6	43.68 5	70.70 11
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMP	73.23 7	56.18 7	59.37 7	54.05 9	47.31 5	61.10 7	58.60 17	42.44 7	71.44 9
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
TAPA-SR		56.09 8	56.78 13	55.62 7	43.28 13	58.11 9	65.17 9	43.60 6	70.29 12

OpenMVS	70.44 10	55.58 9	58.77 8	53.44 11	44.82 10	58.83 8	65.59 8	35.91 17	72.72 7

TAPA-MVS	71.42 9	55.13 10	58.21 10	53.07 12	41.62 16	56.79 11	62.68 12	39.75 12	74.79 3
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
ACMM	72.26 8	55.12 11	54.88 18	55.28 8	38.65 18	61.75 6	60.21 16	43.87 4	71.11 10
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
CMAPSRf		54.99 12	57.09 11	53.58 10	46.78 8	55.24 13	63.74 11	41.76 9	67.41 19

ACMH+	66.54 12	53.99 13	56.90 12	52.05 13	44.66 11	57.37 10	56.79 20	41.98 8	69.14 14

CMAPbl		53.42 14	56.15 14	51.59 14	44.02 12	53.93 14	60.80 15	40.05 11	68.28 17

LTVRE_ROB	59.44 15	53.25 15	56.08 15	51.36 15	39.82 17	53.93 14	62.24 13	37.92 16	72.34 8
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
PLC	68.99 11	51.58 16	55.07 17	49.25 18	41.73 15	52.26 19	56.95 19	38.53 15	68.41 16
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMH	65.37 13	51.50 17	53.46 19	50.20 16	38.46 19	52.73 17	57.24 18	40.63 10	68.46 15
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
GSE		51.31 18	58.55 9	46.48 22	47.08 6	49.58 21	54.32 24	35.53 19	70.02 13

LPCS		50.41 19	55.27 16	47.16 21	43.12 14	51.06 20	54.57 23	35.86 18	67.43 18

COLMAP_ROB	62.73 14	49.91 20	51.76 20	48.68 19	37.30 20	52.31 18	61.83 14	31.91 22	66.23 20
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
cscd		49.00 21	48.08 22	49.61 17	34.83 21	53.10 16	66.82 7	28.93 24	61.32 23

A-TVSNet + Gipuma		45.36 22	47.40 23	44.00 23	29.86 23	43.85 23	55.31 21	32.84 21	64.94 21

CIDER		45.22 23	41.50 24	47.70 20	28.55 24	46.16 22	64.03 10	32.90 20	54.46 25
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
cscdsrtA		41.59 24	48.09 21	37.27 24	33.02 22	27.16 24	55.24 22	29.40 23	63.15 22

PMVS	39.38 17	22.01 25	32.82 25	14.80 25	5.84 28	11.38 26	26.05 25	6.98 27	59.80 24
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
example		18.45 26	28.74 26	11.60 26	6.01 27	5.21 29	22.21 26	7.37 26	51.47 26

MVE	19.12 18	15.60 27	22.59 30	10.94 27	5.22 29	14.12 25	8.87 29	9.84 25	39.95 30
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
DPSNet		15.22 28	24.38 27	9.11 28	6.26 25	5.91 27	17.93 27	3.49 28	42.51 28

hgnet		15.22 28	24.38 27	9.11 28	6.26 25	5.91 27	17.93 27	3.49 28	42.51 28

CMPMVS	47.78 16	9.53 30	23.73 29	0.07 30	0.00 30	0.00 30	0.20 30	0.00 30	47.47 27
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
dnet		0.00 31	0.00 31	0.00 31	0.00 30	0.00 30	0.00 31	0.00 30	0.00 31

Low-resolution many-view benchmark