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.

DeepC-MVS_fast	93.89 2	96.93 1	96.64 2	97.78 2	98.64 2	94.30 2	97.41 9	98.04 6	94.81 2	96.59 2	98.37 4	91.24 2	99.64 7	95.16 3	99.52 1	99.42 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepPCF-MVS	93.97 1	96.61 2	97.09 1	95.15 19	98.09 11	86.63 26	96.00 23	98.15 3	95.43 1	97.95 1	98.56 2	93.40 1	99.36 13	96.77 1	99.48 2	99.45 1

ACMMwithACMP		96.27 3	95.93 3	97.28 5	99.24 1	92.62 6	98.25 4	98.81 1	92.99 8	94.56 5	98.39 3	88.96 5	99.85 1	94.57 4	97.63 14	99.36 4

DeepC-MVS	93.07 3	96.06 4	95.66 4	97.29 4	97.96 13	93.17 4	97.30 11	98.06 5	93.92 5	93.38 7	98.66 1	86.83 9	99.73 4	95.60 2	99.22 4	98.96 7
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
MG-MVS		95.61 5	95.38 5	96.31 12	98.42 6	90.53 15	96.04 22	97.48 13	93.47 7	95.67 3	98.10 6	89.17 4	99.25 16	91.27 10	98.77 6	99.13 6

PHI-MVS		95.53 6	95.15 6	96.66 6	98.32 10	92.42 9	98.11 5	97.72 9	93.53 6	94.13 6	97.28 11	89.22 3	99.34 14	89.83 12	98.20 10	98.22 11

3Dnovator+	91.43 4	95.40 7	94.48 8	98.16 1	96.90 18	95.34 1	98.48 2	97.87 7	94.65 3	88.53 22	98.02 8	83.69 16	99.71 6	93.18 6	98.96 5	99.44 2

3Dnovator	91.36 5	95.19 8	94.44 9	97.44 3	96.56 20	93.36 3	98.65 1	98.36 2	94.12 4	89.25 19	98.06 7	82.20 19	99.77 2	93.41 5	99.32 3	99.18 5

OMC-MVS		95.09 9	94.70 7	96.25 14	98.46 5	91.28 13	96.43 20	97.57 12	92.04 10	94.77 4	97.96 9	87.01 8	99.09 17	91.31 9	96.77 18	98.36 10

AdaColmap		94.34 10	93.68 10	96.31 12	98.59 3	91.68 11	96.59 18	97.81 8	89.87 15	92.15 10	97.06 16	83.62 17	99.54 9	89.34 13	98.07 12	97.70 17

CNLPA		94.28 11	93.53 11	96.52 7	98.38 8	92.55 7	96.59 18	96.88 23	90.13 14	91.91 12	97.24 12	85.21 10	99.09 17	87.64 15	97.83 13	97.92 14

MAR-MVS		94.22 12	93.46 12	96.51 8	98.00 12	92.19 10	97.67 6	97.47 14	88.13 19	93.00 8	95.84 26	84.86 13	99.51 11	87.99 14	98.17 11	97.83 15
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
PLC	91.00 6	94.11 13	93.43 13	96.13 15	98.58 4	91.15 14	96.69 17	97.39 16	87.29 25	91.37 14	96.71 17	88.39 6	99.52 10	87.33 17	97.13 17	97.73 16
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
F-COLMAP		93.58 14	92.98 14	95.37 17	98.40 7	88.98 20	97.18 13	97.29 18	87.75 22	90.49 16	97.10 15	85.21 10	99.50 12	86.70 18	96.72 19	97.63 18

LS3D		93.57 15	92.61 15	96.47 9	97.59 16	91.61 12	97.67 6	97.72 9	85.17 28	90.29 17	98.34 5	84.60 14	99.73 4	83.85 25	98.27 9	98.06 13

CLD-MVS		92.98 16	92.53 16	94.32 22	96.12 26	89.20 19	95.28 25	97.47 14	92.66 9	89.90 18	95.62 28	80.58 20	98.40 27	92.73 7	92.40 24	95.38 24
Zhaoxin Li, Wangmeng Zuo, Zhaoqi Wang, Lei Zhang: Confidence-based Large-scale Dense Multi-view Stereo. IEEE Transaction on Image Processing, 2020
ACMM	89.79 8	92.96 17	92.50 17	94.35 21	96.30 23	88.71 22	97.58 8	97.36 17	91.40 13	90.53 15	96.65 18	79.77 22	98.75 22	91.24 11	91.64 25	95.59 23
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
OpenMVS	89.19 11	92.86 18	91.68 20	96.40 10	95.34 28	92.73 5	98.27 3	98.12 4	84.86 29	85.78 27	97.75 10	78.89 23	99.74 3	87.50 16	98.65 7	96.73 22

ACMP	89.59 9	92.62 19	92.14 18	94.05 24	96.40 22	88.20 23	97.36 10	97.25 20	91.52 11	88.30 23	96.64 19	78.46 26	98.72 23	91.86 8	91.48 26	95.23 26
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
TAPA-MVS	90.10 7	92.30 20	91.22 22	95.56 16	98.33 9	89.60 17	96.79 16	97.65 11	81.83 33	91.52 13	97.23 13	87.94 7	98.91 21	71.31 35	98.37 8	98.17 12
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
PVSNet	86.66 16	92.24 21	91.74 19	93.73 27	97.77 15	83.69 32	92.88 33	96.72 26	87.91 20	93.00 8	94.86 31	78.51 25	99.05 19	86.53 19	97.45 15	98.47 9

PatchmatchNet		91.91 22	91.35 21	93.59 28	95.38 27	84.11 30	93.15 32	95.39 33	89.54 16	92.10 11	93.68 34	82.82 18	98.13 30	84.81 21	95.32 21	98.52 8

PCF-MVS	89.48 10	91.56 23	89.95 25	96.36 11	96.60 19	92.52 8	92.51 34	97.26 19	79.41 34	88.90 20	96.56 20	84.04 15	99.55 8	77.01 32	97.30 16	97.01 20
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
PAPM		91.52 24	90.30 23	95.20 18	95.30 30	89.83 16	93.38 31	96.85 24	86.26 26	88.59 21	95.80 27	84.88 12	98.15 29	75.67 34	95.93 20	97.63 18

MSDG		91.42 25	90.24 24	94.96 20	97.15 17	88.91 21	93.69 30	96.32 28	85.72 27	86.93 25	96.47 21	80.24 21	98.98 20	80.57 30	95.05 22	96.98 21

LTVRE_ROB	88.41 12	90.99 26	89.92 26	94.19 23	96.18 25	89.55 18	96.31 21	97.09 21	87.88 21	85.67 28	95.91 25	78.79 24	98.57 25	81.50 29	89.98 28	94.44 30
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	87.59 15	90.53 27	89.42 27	93.87 25	96.21 24	87.92 24	97.24 12	96.94 22	88.45 18	83.91 31	96.27 23	71.92 28	98.62 24	84.43 22	89.43 31	95.05 27
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
COLMAP_ROB	87.81 14	90.40 28	89.28 28	93.79 26	97.95 14	87.13 25	96.92 15	95.89 31	82.83 32	86.88 26	97.18 14	73.77 27	99.29 15	78.44 31	93.62 23	94.95 28
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
ACMH+	87.92 13	90.20 29	89.18 29	93.25 29	96.48 21	86.45 27	96.99 14	96.68 27	88.83 17	84.79 29	96.22 24	70.16 31	98.53 26	84.42 23	88.04 33	94.77 29

USDC		88.94 30	87.83 30	92.27 31	94.66 33	84.96 29	93.86 29	95.90 30	87.34 24	83.40 32	95.56 29	67.43 32	98.19 28	82.64 27	89.67 30	93.66 32

LF4IMVS		87.94 31	87.25 31	89.98 35	92.38 35	80.05 34	94.38 26	95.25 34	87.59 23	84.34 30	94.74 32	64.31 34	97.66 34	84.83 20	87.45 34	92.23 35

TinyColmap		86.82 32	85.35 32	91.21 32	94.91 32	82.99 33	93.94 28	94.02 37	83.58 31	81.56 34	94.68 33	62.34 35	98.13 30	75.78 33	87.35 35	92.52 33

CMPMVS	62.92 19	85.62 33	84.92 33	87.74 36	89.14 38	73.12 37	94.17 27	96.80 25	73.98 36	73.65 36	94.93 30	66.36 33	97.61 36	83.95 24	91.28 27	92.48 34
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
PVSNet_0	82.17 17	85.46 34	83.64 34	90.92 33	95.27 31	79.49 35	90.55 36	95.60 32	83.76 30	83.00 33	89.95 36	71.09 29	97.97 32	82.75 26	60.79 40	95.31 25

OpenMVS_ROB	81.14 18	84.42 35	82.28 35	90.83 34	90.06 37	84.05 31	95.73 24	94.04 36	73.89 37	80.17 35	91.53 35	59.15 36	97.64 35	66.92 36	89.05 32	90.80 36

FPMVS		71.27 36	69.85 36	75.50 38	74.64 39	59.03 41	91.30 35	91.50 38	58.80 39	57.92 38	88.28 37	29.98 39	85.53 39	53.43 39	82.84 36	81.95 38

Gipuma		67.86 37	65.41 37	75.18 39	92.66 34	73.45 36	66.50 42	94.52 35	53.33 40	57.80 39	66.07 42	30.81 38	89.20 37	48.15 42	78.88 37	62.90 44
S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion. ICCV 2015
ANet_high		63.94 38	59.58 38	77.02 37	61.24 44	66.06 38	85.66 37	87.93 39	78.53 35	42.94 40	71.04 40	25.42 40	80.71 41	52.60 40	30.83 44	84.28 37

PNet_i23d		59.01 39	55.87 39	68.44 42	73.98 40	51.37 43	81.36 38	82.41 42	52.37 41	42.49 41	70.39 41	11.39 41	79.99 43	49.77 41	38.71 42	73.97 41

PMVS	53.92 20	58.58 40	55.40 40	68.12 43	51.00 45	48.64 44	78.86 40	87.10 40	46.77 43	35.84 43	74.28 39	8.76 42	86.34 38	42.07 43	73.91 38	69.38 42
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
wuykxyi23d		56.92 41	51.11 41	74.38 41	62.30 43	61.47 40	80.09 39	84.87 41	49.62 42	30.80 44	57.20 44	7.03 43	82.94 40	55.69 38	32.36 43	78.72 40

MVE	50.73 21	53.25 42	48.81 42	66.58 44	65.34 42	57.50 42	72.49 41	70.94 44	40.15 44	39.28 42	63.51 43	6.89 45	73.48 44	38.29 44	42.38 41	68.76 43
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
wuyk23d		25.11 43	24.57 43	26.74 45	73.98 40	39.89 45	57.88 43	9.80 45	12.27 45	10.39 45	6.97 45	7.03 43	36.44 45	25.43 45	17.39 45	3.89 45

DeepMVS_CX				74.68 40	90.84 36	64.34 39		81.61 43	65.34 38	67.47 37	88.01 38	48.60 37	80.13 42	62.33 37	73.68 39	79.58 39

ITE_SJBPF				92.43 30	95.34 28	85.37 28		95.92 29	91.47 12	87.75 24	96.39 22	71.00 30	97.96 33	82.36 28	89.86 29	93.97 31

High-resolution multi-view benchmark