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	botani.	boulde.	bridge	door	exhibi.	lectur.	living.	lounge	observ.	old co.	statue	terrac.

LS3D		99.27 1	99.12 1	99.74 1	99.18 7	99.75 1	99.56 3	99.57 3	98.45 2	99.49 1	99.85 3	97.77 2	99.94 1	98.33 4	99.84 3	99.52 2

MSDG		98.98 2	98.80 2	99.53 2	99.76 2	99.19 2	98.75 7	99.55 5	97.25 6	99.47 2	99.77 5	97.82 1	99.87 3	96.93 7	99.90 1	99.54 1

COLMAP_ROB	97.56 1	98.86 3	98.75 3	99.17 3	99.88 1	98.53 3	99.34 5	99.59 2	97.55 5	98.70 5	99.89 2	95.83 3	99.90 2	98.10 6	99.90 1	99.08 3
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
ACMM		98.37 4	98.34 4	98.48 4	99.41 3	97.10 4	99.56 3	99.45 7	98.53 1	99.04 4	99.85 3	93.00 4	99.71 5	98.74 2	97.45 6	98.64 6

MVELAS		98.16 5	98.10 5	98.33 7	99.29 5	96.82 6	98.75 7	99.44 9	97.83 4	99.13 3	99.55 8	92.92 5	99.67 7	98.32 5	97.69 5	98.48 7

ACMH	97.28 2	98.10 6	97.99 6	98.44 5	99.41 3	96.96 5	99.60 2	99.56 4	98.09 3	98.15 7	99.91 1	90.87 7	99.70 6	98.88 1	97.45 6	98.67 5

LTVRE_ROB	97.16 3	98.02 7	97.90 7	98.40 6	99.23 6	96.80 7	99.70 1	99.60 1	97.12 7	98.18 6	99.70 6	91.73 6	99.72 4	98.39 3	97.45 6	98.68 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
OpenMVS_ROB	92.34 4	94.38 8	93.70 9	96.41 8	97.38 9	93.17 8	99.06 6	98.75 11	86.58 9	94.84 8	98.26 10	81.53 9	99.32 8	89.01 9	97.87 4	96.76 8

CMPMVS	69.68 7	94.13 9	94.90 8	91.84 10	97.24 10	80.01 10	98.52 9	99.48 6	89.01 8	91.99 10	99.67 7	85.67 8	99.13 9	95.44 8	97.03 9	96.39 9
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
Gipuma		90.99 10	90.15 10	93.51 9	98.73 8	90.12 9	93.98 11	99.45 7	79.32 10	92.28 9	94.91 11	69.61 10	97.98 10	87.42 10	95.67 10	92.45 11
S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion. ICCV 2015
MVE	76.82 5	76.91 11	74.31 12	84.70 11	85.38 12	76.05 11	96.88 10	93.17 13	67.39 12	71.28 12	89.01 13	21.66 13	87.69 13	71.74 13	72.29 13	90.35 12
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
PMVS	70.75 6	75.98 12	74.97 11	79.01 13	70.98 13	55.18 13	93.37 12	98.21 12	65.08 13	61.78 13	93.83 12	21.74 12	92.53 11	78.59 12	91.12 12	89.34 13
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
FPMVS				82.75 12	86.77 11	63.39 12		98.92 10	74.11 11	83.39 11	98.98 9	50.85 11	92.40 12	84.54 11	94.97 11	92.46 10

High-resolution multi-view benchmark