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	89.96 1	94.20 1	94.77 1	92.49 13	96.52 11	80.00 30	94.00 26	97.08 4	90.05 2	95.65 1	97.29 2	89.66 1	98.97 11	93.95 1	98.71 2	98.50 2

DeepC-MVS_fast	89.43 2	94.04 2	93.79 2	94.80 2	97.48 3	86.78 2	95.65 11	96.89 7	89.40 4	92.81 3	96.97 4	85.37 4	99.24 5	90.87 4	98.69 3	98.38 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
PHI-MVS		93.89 3	93.65 3	94.62 3	96.84 8	86.43 3	96.69 6	97.49 1	85.15 16	93.56 2	96.28 8	85.60 3	99.31 4	92.45 2	98.79 1	98.12 5

DeepC-MVS	88.79 3	93.31 4	92.99 5	94.26 6	96.07 15	85.83 5	94.89 17	96.99 6	89.02 6	89.56 8	97.37 1	82.51 6	99.38 2	92.20 3	98.30 4	97.57 11
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMMP		93.24 5	92.88 6	94.30 5	98.09 1	85.33 7	96.86 5	97.45 2	88.33 7	90.15 5	97.03 3	81.44 8	99.51 1	90.85 5	95.74 14	98.04 7

CSCG		93.23 6	93.05 4	93.76 7	98.04 2	84.07 12	96.22 8	97.37 3	84.15 17	90.05 6	95.66 13	87.77 2	99.15 8	89.91 6	98.27 5	98.07 6

3Dnovator+	87.14 4	92.42 7	91.37 9	95.55 1	95.63 21	88.73 1	97.07 3	96.77 8	90.84 1	84.02 27	96.62 6	75.95 20	99.34 3	87.77 7	97.68 7	98.59 1

MG-MVS		91.77 8	91.70 7	92.00 20	97.08 6	80.03 28	93.60 27	95.18 23	87.85 9	90.89 4	96.47 7	82.06 7	98.36 18	85.07 14	97.04 8	97.62 10

3Dnovator	86.66 5	91.73 9	90.82 11	94.44 4	94.59 28	86.37 4	97.18 2	97.02 5	89.20 5	84.31 25	96.66 5	73.74 24	99.17 7	86.74 11	97.96 6	97.79 9

EPP-MVSNet		91.70 10	91.56 8	92.13 19	95.88 17	80.50 26	97.33 1	95.25 21	86.15 13	89.76 7	95.60 14	83.42 5	98.32 20	87.37 8	93.25 25	97.56 12

IS-MVSNet		91.43 11	91.09 10	92.46 14	95.87 19	81.38 20	96.95 4	93.69 30	89.72 3	89.50 10	95.98 11	78.57 13	97.77 26	83.02 15	96.50 13	98.22 4

OMC-MVS		91.23 12	90.62 12	93.08 8	96.27 14	84.07 12	93.52 28	95.93 13	86.95 11	89.51 9	96.13 9	78.50 14	98.35 19	85.84 12	92.90 27	96.83 14

API-MVS		90.66 13	90.07 13	92.45 15	96.36 12	84.57 9	96.06 9	95.22 22	82.39 22	89.13 11	94.27 24	80.32 10	98.46 16	80.16 20	96.71 10	94.33 30

MAR-MVS		90.30 14	89.37 15	93.07 10	96.61 10	84.48 10	95.68 10	95.67 17	82.36 23	87.85 12	92.85 29	76.63 19	98.80 13	80.01 21	96.68 11	95.91 18
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
AdaColmap		89.89 15	89.07 16	92.37 16	97.41 4	83.03 14	94.42 21	95.92 14	82.81 21	86.34 15	94.65 19	73.89 23	99.02 9	80.69 19	95.51 15	95.05 23

Vis-MVSNet (Re-imp)		89.59 16	89.44 14	90.03 33	95.74 20	75.85 39	95.61 12	90.80 45	87.66 10	87.83 13	95.40 16	76.79 17	96.46 39	78.37 23	96.73 9	97.80 8

CLD-MVS		89.47 17	88.90 17	91.18 25	94.22 31	82.07 18	92.13 33	96.09 12	87.90 8	85.37 18	92.45 33	74.38 22	97.56 29	87.15 9	90.43 29	93.93 32
Zhaoxin Li, Wangmeng Zuo, Zhaoqi Wang, Lei Zhang: Confidence-based Large-scale Dense Multi-view Stereo. IEEE Transaction on Image Processing, 2020
PAPR		89.15 18	88.32 20	91.64 21	95.50 22	80.01 29	90.64 36	95.41 20	79.91 29	84.05 26	92.10 34	81.08 9	97.74 27	81.36 17	94.79 18	97.17 13

ACMM	84.12 9	89.14 19	88.48 18	91.12 26	94.65 27	81.22 22	95.31 14	96.12 11	85.31 15	85.92 17	94.34 21	70.19 30	98.06 22	85.65 13	88.86 31	94.08 31
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
CNLPA		89.07 20	87.98 22	92.34 17	96.87 7	84.78 8	94.08 25	93.24 32	81.41 27	84.46 22	95.13 18	75.57 21	96.62 37	77.21 27	93.84 21	95.61 20

PLC	84.53 7	89.06 21	88.03 21	92.15 18	97.27 5	82.69 17	94.29 23	95.44 19	79.71 31	84.01 28	94.18 25	76.68 18	98.75 14	77.28 26	93.41 24	95.02 24
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMP	84.23 8	89.01 22	88.35 19	90.99 27	94.73 24	81.27 21	95.07 16	95.89 15	86.48 12	83.67 29	94.30 23	69.33 33	97.99 23	87.10 10	88.55 32	93.72 33
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
OpenMVS	83.78 11	88.74 23	87.29 23	93.08 8	92.70 37	85.39 6	96.57 7	96.43 9	78.74 36	80.85 34	96.07 10	69.64 32	99.01 10	78.01 24	96.65 12	94.83 26

TAPA-MVS	84.62 6	88.16 24	87.01 24	91.62 22	96.64 9	80.65 24	94.39 22	96.21 10	76.38 39	86.19 16	95.44 15	79.75 12	98.08 21	62.75 48	95.29 16	96.13 15
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
F-COLMAP		87.95 25	86.80 25	91.40 23	96.35 13	80.88 23	94.73 18	95.45 18	79.65 32	82.04 32	94.61 20	71.13 29	98.50 15	76.24 29	91.05 28	94.80 28

LS3D		87.89 26	86.32 26	92.59 12	96.07 15	82.92 15	95.23 15	94.92 24	75.66 42	82.89 30	95.98 11	72.48 26	99.21 6	68.43 44	95.23 17	95.64 19

PCF-MVS	84.11 10	87.74 27	86.08 27	92.70 11	94.02 32	84.43 11	89.27 42	95.87 16	73.62 46	84.43 23	94.33 22	78.48 15	98.86 12	70.27 40	94.45 19	94.81 27
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
PatchMatch-RL		86.77 28	85.54 28	90.47 29	95.88 17	82.71 16	90.54 37	92.31 38	79.82 30	84.32 24	91.57 38	68.77 34	96.39 41	73.16 35	93.48 23	92.32 38

PAPM		86.68 29	85.39 29	90.53 28	93.05 35	79.33 34	89.79 39	94.77 27	78.82 34	81.95 33	93.24 28	76.81 16	97.30 31	66.94 45	93.16 26	94.95 25

LTVRE_ROB	82.13 12	86.26 30	84.90 31	90.34 31	94.44 30	81.50 19	92.31 32	94.89 25	83.03 19	79.63 37	92.67 31	69.69 31	97.79 25	71.20 39	86.26 36	91.72 39
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
PatchmatchNet		85.85 31	84.70 32	89.29 36	91.76 40	75.54 40	88.49 47	91.30 42	81.63 26	85.05 19	88.70 46	71.71 28	96.24 43	74.61 32	89.05 30	96.08 16
Fangjinhua Wang, Silvano Galliani, Christoph Vogel, Pablo Speciale, Marc Pollefeys: PatchmatchNet: Learned Multi-View Patchmatch Stereo.
PMMVS		85.71 32	84.96 30	87.95 39	88.90 46	77.09 36	88.68 45	90.06 49	72.32 49	86.47 14	90.76 41	72.15 27	94.40 48	81.78 16	93.49 22	92.36 37

PVSNet	78.82 17	85.55 33	84.65 33	88.23 38	94.72 25	71.93 46	87.12 48	92.75 36	78.80 35	84.95 21	90.53 42	64.43 37	96.71 35	74.74 31	93.86 20	96.06 17

ACMH	80.38 16	85.36 34	83.68 35	90.39 30	94.45 29	80.63 25	94.73 18	94.85 26	82.09 25	77.24 42	92.65 32	60.01 41	97.58 28	72.25 37	84.87 40	92.96 35
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
IB-MVS	80.51 14	85.24 35	83.26 38	91.19 24	92.13 38	79.86 31	91.75 34	91.29 43	83.28 18	80.66 35	88.49 47	61.28 39	98.46 16	80.99 18	79.46 47	95.25 22

RPSCF		85.07 36	84.27 34	87.48 42	92.91 36	70.62 48	91.69 35	92.46 37	76.20 41	82.67 31	95.22 17	63.94 38	97.29 32	77.51 25	85.80 38	94.53 29

ACMH+	81.04 13	85.05 37	83.46 37	89.82 34	94.66 26	79.37 33	94.44 20	94.12 28	82.19 24	78.04 40	92.82 30	58.23 44	97.54 30	73.77 34	82.90 44	92.54 36

MSDG		84.86 38	83.09 39	90.14 32	93.80 33	80.05 27	89.18 44	93.09 34	78.89 33	78.19 39	91.91 36	65.86 36	97.27 33	68.47 43	88.45 33	93.11 34

TAMVS		84.49 39	83.59 36	87.19 45	89.00 45	73.61 43	95.60 13	93.11 33	82.97 20	78.87 38	91.22 39	72.80 25	96.75 34	78.58 22	70.12 50	91.19 42

COLMAP_ROB	80.39 15	83.96 40	82.04 41	89.74 35	95.28 23	79.75 32	94.25 24	92.28 39	75.17 44	78.02 41	93.77 26	58.60 43	97.84 24	65.06 46	85.92 37	91.63 40
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
EPMVS		83.90 41	82.70 40	87.51 41	90.23 44	72.67 45	88.62 46	81.96 53	81.37 28	85.01 20	88.34 48	66.31 35	94.45 47	75.30 30	87.12 35	95.43 21

USDC		82.76 42	81.26 42	87.26 43	91.17 42	74.55 42	89.27 42	93.39 31	78.26 37	75.30 44	92.08 35	54.43 47	96.63 36	71.64 38	85.79 39	90.61 44

test_0402		81.30 43	79.17 44	87.67 40	93.19 34	78.17 35	92.98 30	91.71 41	75.25 43	76.02 43	90.31 43	59.23 42	96.37 42	50.22 50	83.63 43	88.47 48

CMPMVS	59.16 20	80.52 44	79.20 43	84.48 48	83.98 50	67.63 50	89.95 38	93.84 29	64.79 51	66.81 50	91.14 40	57.93 45	95.17 46	76.25 28	88.10 34	90.65 43
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
LF4IMVS		80.37 45	79.07 45	84.27 50	86.64 49	69.87 49	89.39 41	91.05 44	76.38 39	74.97 46	90.00 45	47.85 50	94.25 49	74.55 33	80.82 46	88.69 46

TDRefinement		79.81 46	77.34 47	87.22 44	79.24 51	75.48 41	93.12 29	92.03 40	76.45 38	75.01 45	91.58 37	49.19 49	96.44 40	70.22 41	69.18 51	89.75 45

TinyColmap		79.76 47	77.69 46	85.97 47	91.71 41	73.12 44	89.55 40	90.36 47	75.03 45	72.03 49	90.19 44	46.22 51	96.19 45	63.11 47	81.03 45	88.59 47

OpenMVS_ROB	74.94 18	79.51 48	77.03 48	86.93 46	87.00 48	76.23 38	92.33 31	90.74 46	68.93 50	74.52 47	88.23 49	49.58 48	96.62 37	57.64 49	84.29 42	87.94 50

PVSNet_0	73.20 19	77.22 49	74.83 49	84.37 49	90.70 43	71.10 47	83.09 50	89.67 50	72.81 48	73.93 48	83.13 50	60.79 40	93.70 50	68.54 42	50.84 54	88.30 49

FPMVS		64.63 50	62.55 50	70.88 52	70.80 53	56.71 52	84.42 49	84.42 52	51.78 52	49.57 51	81.61 51	23.49 53	81.48 53	40.61 54	76.25 48	74.46 53

ANet_high		58.88 51	54.22 52	72.86 51	56.50 58	56.67 53	80.75 51	86.00 51	73.09 47	37.39 54	64.63 54	22.17 54	79.49 55	43.51 52	23.96 58	82.43 51

Gipuma		57.99 52	54.91 51	67.24 54	88.51 47	65.59 51	52.21 56	90.33 48	43.58 55	42.84 53	51.18 57	20.29 55	85.07 51	34.77 56	70.45 49	51.05 58
S. Galliani, K. Lasinger, K. Schindler: Massively Parallel Multiview Stereopsis by Surface Normal Diffusion. ICCV 2015
PMVS	47.18 21	52.22 53	48.46 53	63.48 55	45.72 59	46.20 57	73.41 52	78.31 55	41.03 56	30.06 56	65.68 53	6.05 57	83.43 52	30.04 57	65.86 52	60.80 55
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
wuykxyi23d		50.55 54	44.13 55	69.81 53	56.77 56	54.58 54	73.22 53	80.78 54	39.79 57	22.08 58	46.69 58	4.03 59	79.71 54	47.65 51	26.13 57	75.14 52

PNet_i23d		50.48 55	47.18 54	60.36 56	68.59 54	44.56 58	72.75 54	72.61 56	43.92 54	33.91 55	60.19 55	6.16 56	73.52 56	38.50 55	28.04 56	63.01 54

MVE	39.65 22	43.39 56	38.59 56	57.77 57	56.52 57	48.77 56	55.38 55	58.64 57	29.33 58	28.96 57	52.65 56	4.68 58	64.62 57	28.11 58	33.07 55	59.93 56
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
wuyk23d		21.27 57	20.48 57	23.63 59	68.59 54	36.41 59	49.57 57	6.85 59	9.37 59	7.89 59	4.46 59	4.03 59	31.37 59	17.47 59	16.07 59	3.12 59

Test By Simon												80.02 11

ITE_SJBPF				88.24 37	91.88 39	77.05 37		92.92 35	85.54 14	80.13 36	93.30 27	57.29 46	96.20 44	72.46 36	84.71 41	91.49 41

DeepMVS_CX				56.31 58	74.23 52	51.81 55		56.67 58	44.85 53	48.54 52	75.16 52	27.87 52	58.74 58	40.92 53	52.22 53	58.39 57

High-resolution multi-view benchmark