High-res multi-view results

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.

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	high-res multi-view	indoor	outdoor	courty.	delive.	electro	facade	kicker	meadow	office	pipes	playgr.	relief	relief.	terrace	terrai.

MVELAS		79.35 1	81.23 1	77.16 2	85.01 3	86.92 1	85.87 2	60.89 6	80.07 1	75.35 1	72.96 1	73.21 1	68.43 4	85.41 1	84.63 1	87.41 2	85.44 2

ACMM		78.86 2	78.13 2	79.71 1	86.89 1	83.40 2	86.02 1	70.50 1	75.28 2	71.49 2	63.01 2	69.26 2	73.57 1	84.11 2	83.98 2	89.76 1	87.84 1

LS3D		74.08 3	73.39 3	74.88 3	85.05 2	82.62 3	79.71 3	68.66 2	72.82 3	58.80 4	57.61 3	61.31 3	71.07 2	80.32 4	78.87 4	86.00 5	80.18 5

MSDG		71.52 4	69.87 5	73.44 4	82.21 5	79.35 5	79.52 4	64.59 4	66.15 4	61.87 3	53.21 5	56.09 4	65.85 5	78.94 5	78.50 5	86.60 3	76.85 6

ACMH	65.37 1	70.71 5	70.00 4	71.54 5	82.51 4	82.47 4	77.78 5	68.13 3	56.19 6	46.06 7	54.30 4	51.20 6	68.68 3	80.66 3	80.72 3	86.07 4	84.45 3

COLMAP_ROB	62.73 2	67.66 6	66.76 6	68.70 6	80.49 6	77.98 6	75.29 6	62.95 5	63.62 5	49.96 6	47.32 7	50.72 7	58.57 7	76.87 6	75.50 7	84.94 6	75.33 8
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
CMPMVS	47.78 4	62.49 7	62.52 8	62.46 7	70.01 8	70.66 8	62.97 8	51.84 9	51.98 7	56.71 5	42.87 8	53.62 5	57.80 8	72.23 8	70.37 8	75.45 8	75.91 7
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
LTVRE_ROB	59.44 3	61.82 8	62.64 7	60.87 8	72.83 7	77.19 7	64.37 7	58.97 7	33.56 10	28.00 10	52.59 6	42.21 8	63.93 6	74.52 7	76.28 6	77.15 7	82.13 4
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
FPMVS		51.43 9	49.18 9	54.07 9	66.83 9	57.21 9	60.25 9	50.91 10	49.58 8	34.36 9	31.02 9	32.02 9	41.49 9	58.98 10	56.07 10	70.56 9	59.36 9

PMVS	39.38 5	46.06 10	43.30 10	49.28 10	62.93 10	38.75 10	41.88 10	53.50 8	33.33 11	35.46 8	28.90 10	31.01 10	33.04 10	58.61 11	54.63 11	68.86 10	57.88 10
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
Gipuma		36.38 11	35.80 11	37.07 11	45.76 11	33.90 11	29.81 11	48.47 11	39.91 9	18.02 11	8.00 12	8.14 11	25.14 11	59.29 9	61.02 9	55.19 11	40.31 11
S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion. ICCV 2015
MVE	19.12 6	20.47 12	23.27 12	17.20 12	12.66 12	25.41 12	10.52 12	34.14 12	14.79 12	6.53 12	8.79 11	4.68 12	16.64 12	29.49 12	41.63 12	22.73 12	38.11 12
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)

High-resolution multi-view benchmark