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.

DeepPCF-MVS	69.58 1	79.03 1	79.00 1	79.13 3	84.92 5	60.32 4	83.03 7	85.33 3	62.86 4	80.17 1	90.03 2	61.76 1	88.95 4	74.21 2	88.67 2	88.12 2

DeepC-MVS	69.38 2	78.56 2	78.14 2	79.83 1	83.60 6	61.62 2	84.17 4	86.85 1	63.23 2	73.84 2	90.25 1	57.68 2	89.96 1	74.62 1	89.03 1	87.89 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepC-MVS_fast	68.24 3	77.25 3	76.63 3	79.12 4	86.15 1	60.86 3	84.71 3	84.85 4	61.98 6	73.06 3	88.88 5	53.72 3	89.06 3	68.27 4	88.04 3	87.42 5
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMMwithACMP		76.02 4	75.33 4	78.07 5	85.20 3	61.91 1	85.49 2	84.44 6	63.04 3	69.80 6	89.74 3	45.43 12	87.16 8	72.01 3	82.87 11	85.14 8

3Dnovator+	66.72 4	75.84 5	74.57 5	79.66 2	82.40 7	59.92 5	85.83 1	86.32 2	66.92 1	67.80 11	89.24 4	42.03 15	89.38 2	64.07 7	86.50 4	89.69 1

MG-MVS		73.96 6	73.89 6	74.16 12	85.65 2	49.69 25	81.59 11	81.29 13	61.45 7	71.05 4	88.11 6	51.77 6	87.73 6	61.05 9	83.09 9	85.05 10

CLD-MVS		73.33 7	72.68 7	75.29 10	78.82 15	53.33 19	78.23 21	84.79 5	61.30 8	70.41 5	81.04 24	52.41 5	87.12 9	64.61 6	82.49 12	85.41 7
Zhaoxin Li, Wangmeng Zuo, Zhaoqi Wang, Lei Zhang: Confidence-based Large-scale Dense Multi-view Stereo. IEEE Transaction on Image Processing, 2020
3Dnovator	64.47 5	72.49 8	71.39 9	75.79 6	77.70 20	58.99 7	80.66 14	83.15 7	62.24 5	65.46 14	86.59 7	42.38 14	85.52 12	59.59 11	84.72 5	82.85 12

ACMP	63.53 6	72.30 9	71.20 10	75.59 8	80.28 13	57.54 11	82.74 8	82.84 9	60.58 9	65.24 15	86.18 9	39.25 17	86.03 11	66.95 5	76.79 18	83.22 11
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
0108		72.17 10	71.41 8	74.45 11	81.95 9	57.22 12	84.03 5	80.38 15	59.89 10	68.40 9	82.33 20	49.64 8	87.83 5	51.87 12	84.16 7	78.30 20

MAR-MVS		71.51 11	70.15 13	75.60 7	81.84 10	59.39 6	81.38 12	82.90 8	54.90 17	68.08 10	78.70 27	47.73 10	85.51 13	51.68 14	84.17 6	81.88 13
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
OMC-MVS		71.40 12	70.60 11	73.78 15	76.60 24	53.15 20	79.74 17	79.78 18	58.37 12	68.75 8	86.45 8	45.43 12	80.60 23	62.58 8	77.73 16	87.58 4

PHI-MVS		71.29 13	70.59 12	73.38 16	82.12 8	57.05 13	83.92 6	79.92 17	59.44 11	67.80 11	77.55 31	50.20 7	87.36 7	50.55 15	83.85 8	75.74 28

PCF-MVS	61.88 8	70.95 14	69.49 15	75.35 9	77.63 21	55.71 14	76.04 26	81.81 11	50.30 24	69.66 7	85.40 11	52.51 4	84.89 15	51.82 13	80.24 13	85.45 6
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMM	61.98 7	70.80 15	69.73 14	74.02 13	80.59 12	58.59 8	82.68 9	82.02 10	55.46 16	67.18 13	84.39 14	38.51 20	83.17 18	60.65 10	76.10 19	80.30 15
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
AdaColmap		69.99 16	68.66 16	73.97 14	84.94 4	57.83 10	82.63 10	78.71 20	56.28 14	64.34 17	84.14 15	41.57 16	87.06 10	46.45 18	78.88 15	77.02 25

OpenMVS	61.03 9	68.85 17	67.56 17	72.70 17	74.26 28	53.99 17	81.21 13	81.34 12	52.70 19	62.75 18	85.55 10	38.86 19	84.14 17	48.41 17	83.01 10	79.97 17

PAPM		67.92 18	66.69 18	71.63 19	78.09 17	49.02 26	77.09 24	81.24 14	51.04 22	60.91 20	83.98 16	47.71 11	84.99 14	40.81 25	79.32 14	80.90 14

TAPA-MVS	59.36 10	66.60 19	65.20 19	70.81 21	76.63 23	48.75 27	76.52 25	80.04 16	50.64 23	65.24 15	84.93 12	39.15 18	78.54 27	36.77 30	76.88 17	85.14 8
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
ACMH+	57.40 11	66.12 20	64.06 20	72.30 18	77.79 19	52.83 21	80.39 15	78.03 21	57.30 13	57.47 23	82.55 18	27.68 30	84.17 16	45.54 19	69.78 27	79.90 18

CNLPA		65.43 21	64.02 21	69.68 24	78.73 16	58.07 9	77.82 23	70.71 30	51.49 20	61.57 19	83.58 17	38.23 21	70.82 32	43.90 20	70.10 26	80.16 16

ACMH	55.70 13	65.20 22	63.57 22	70.07 23	78.07 18	52.01 24	79.48 18	79.69 19	55.75 15	56.59 26	80.98 25	27.12 31	80.94 22	42.90 22	71.58 23	77.25 24
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PLC	56.13 12	65.09 23	63.21 23	70.72 22	81.04 11	54.87 16	78.57 19	77.47 22	48.51 26	55.71 28	81.89 21	33.71 23	79.71 24	41.66 24	70.37 25	77.58 22
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
LS3D		64.71 24	62.50 24	71.34 20	79.72 14	55.71 14	79.82 16	74.72 26	48.50 27	56.62 25	84.62 13	33.59 24	82.34 19	29.65 37	75.23 20	75.97 27

LTVRE_ROB	55.42 14	63.15 25	61.23 25	68.92 27	76.57 25	47.80 28	59.92 35	76.39 23	54.35 18	58.67 22	82.46 19	29.44 28	81.49 20	42.12 23	71.14 24	77.46 23
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
F-COLMAP		63.05 26	60.87 26	69.58 25	76.99 22	53.63 18	78.12 22	76.16 24	47.97 28	52.41 31	81.61 22	27.87 29	78.11 28	40.07 26	66.66 29	77.00 26

MSDG		61.81 27	59.23 27	69.55 26	72.64 30	52.63 22	70.45 29	75.81 25	51.38 21	53.70 30	76.11 32	29.52 26	81.08 21	37.70 29	65.79 31	74.93 30

COLMAP_ROB	52.97 15	61.27 28	58.81 28	68.64 28	74.63 27	52.51 23	78.42 20	73.30 28	49.92 25	50.96 32	81.51 23	23.06 34	79.40 26	31.63 35	65.85 30	74.01 31
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
OpenMVS_ROB	52.78 16	60.03 29	58.14 30	65.69 29	70.47 31	44.82 32	75.33 27	70.86 29	45.04 33	56.06 27	76.00 33	26.89 33	79.65 25	35.36 32	67.29 28	72.60 33

PatchmatchNet		59.84 30	58.24 29	64.65 30	73.05 29	46.70 31	69.42 30	62.18 38	47.55 30	58.88 21	71.96 36	34.49 22	69.16 33	42.99 21	63.60 32	78.07 21

PVSNet	50.76 17	58.40 31	57.39 31	61.42 35	75.53 26	44.04 35	61.43 34	63.45 37	47.04 32	56.91 24	73.61 35	27.00 32	64.76 35	39.12 27	72.40 22	75.47 29

CMPMVS	42.80 19	57.81 32	55.97 32	63.32 31	60.98 38	47.38 29	64.66 33	69.50 32	32.06 38	46.83 34	77.80 29	29.50 27	71.36 31	48.68 16	73.75 21	71.21 36
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
USDC		56.35 33	54.24 33	62.69 32	64.74 35	40.31 38	65.05 32	73.83 27	43.93 35	47.58 33	77.71 30	15.36 36	75.05 29	38.19 28	61.81 34	72.70 32

TinyColmap		54.14 34	51.72 34	61.40 36	66.84 33	41.97 36	66.52 31	68.51 33	44.81 34	42.69 37	75.77 34	11.66 39	72.94 30	31.96 34	56.77 35	69.27 37

TDRefinement		53.44 35	50.72 35	61.60 34	64.31 36	46.96 30	70.89 28	65.27 35	41.78 36	44.61 36	77.98 28	11.52 40	66.36 34	28.57 38	51.59 37	71.49 35

PVSNet_0	43.31 18	47.46 36	45.64 36	52.92 38	67.60 32	44.65 33	54.06 39	54.64 39	41.59 37	46.15 35	58.75 39	30.99 25	58.66 40	32.18 33	24.81 41	55.46 39

LF4IMVS		42.95 37	42.26 37	45.04 40	48.30 40	32.50 41	54.80 37	48.49 42	28.03 40	40.51 38	70.16 37	9.24 41	43.89 43	31.63 35	49.18 39	58.72 38

FPMVS		42.18 38	41.11 38	45.39 39	58.03 39	41.01 37	49.50 40	53.81 40	30.07 39	33.71 39	64.03 38	11.69 38	52.08 42	14.01 43	55.11 36	43.09 41

ANet_high		41.38 39	37.47 39	53.11 37	39.73 43	24.45 46	56.94 36	69.69 31	47.65 29	26.04 40	52.32 40	12.44 37	62.38 38	21.80 40	10.61 46	72.49 34

PMVS	28.69 20	36.22 40	33.29 40	45.02 41	36.82 44	35.98 40	54.68 38	48.74 41	26.31 41	21.02 41	51.61 41	2.88 44	60.10 39	9.99 45	47.58 40	38.99 43
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
Gipuma		34.77 41	31.91 41	43.33 43	62.05 37	37.87 39	20.39 44	67.03 34	23.23 42	18.41 42	25.84 45	4.24 42	62.73 36	14.71 42	51.32 38	29.38 45
S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion. ICCV 2015
wuykxyi23d		28.12 42	22.54 43	44.87 42	34.97 45	32.11 42	37.96 42	47.31 43	13.32 44	9.29 46	23.72 46	0.45 47	56.58 41	21.85 39	13.98 43	45.93 40

PNet_i23d		27.88 43	25.99 42	33.55 44	47.54 41	25.89 44	47.24 41	32.91 44	21.44 43	15.90 43	38.09 42	0.85 46	42.76 44	16.90 41	13.03 44	32.00 44

MVE	17.77 21	21.41 44	17.77 44	32.34 45	34.34 46	25.44 45	16.11 45	24.11 45	11.19 45	13.22 44	31.92 43	1.58 45	30.95 45	10.47 44	17.03 42	40.62 42
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
wuyk23d		13.32 45	12.52 45	15.71 46	47.54 41	26.27 43	31.06 43	1.98 47	4.93 47	5.18 47	1.94 47	0.45 47	18.54 46	6.81 47	12.83 45	2.33 47

Test By Simon												48.33 9

DeepMVS_CX				12.03 47	17.97 47	10.91 47		10.60 46	7.46 46	11.07 45	28.36 44	3.28 43	11.29 47	8.01 46	9.74 47	13.89 46

ITE_SJBPF				62.09 33	66.16 34	44.55 34		64.32 36	47.36 31	55.31 29	80.34 26	19.27 35	62.68 37	36.29 31	62.39 33	79.04 19

High-resolution multi-view benchmark