Low-res many-view results

This table lists the benchmark results for the low-res many-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	low-res many-view	indoor	outdoor	lakeside	sand box	storage room	storage room 2	tunnel

A1Net		62.59 1	54.91 3	67.70 6	68.95 3	72.10 7	52.15 2	57.67 3	62.06 9

ACMP	71.68 7	62.33 2	51.03 7	69.86 3	66.41 5	74.93 1	47.36 8	54.70 7	68.24 1
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
A-TVSNet + Gipuma		61.02 3	56.34 1	64.14 10	64.32 7	66.91 18	52.09 3	60.60 1	61.19 13

COLMAP(SR)		60.84 4	47.20 10	69.93 2	69.97 2	72.99 4	47.93 6	46.47 13	66.85 2

DeepPCF-MVS	81.17 1	60.64 5	53.51 5	65.40 8	64.27 8	69.40 12	51.28 5	55.74 6	62.53 7

ACMH+	65.35 13	59.83 6	46.18 13	68.93 4	65.70 6	74.24 3	46.34 12	46.01 14	66.83 3

PVSNet	68.08 12	59.14 7	42.70 16	70.10 1	74.33 1	74.52 2	38.12 18	47.28 12	61.46 11

TAPA-MVS	70.22 9	58.89 8	51.21 6	64.01 11	62.59 11	66.45 20	46.31 13	56.11 5	62.99 6
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
DeepC-MVS	77.92 3	58.46 9	50.82 8	63.55 13	62.52 12	66.82 19	47.58 7	54.06 8	61.30 12

PCF-MVS	73.15 6	58.42 10	54.11 4	61.29 16	61.25 15	64.94 22	51.80 4	56.43 4	57.66 16
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMM	69.62 10	58.27 11	46.31 12	66.24 7	60.67 17	72.46 5	42.63 15	50.00 10	65.60 5
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
DeepC-MVS_fast	78.06 2	58.14 12	48.66 9	64.46 9	63.44 9	67.77 15	46.93 10	50.38 9	62.16 8

TAPA-MVS(SR)		55.57 13	35.84 25	68.73 5	67.52 4	72.24 6	33.19 26	38.50 22	66.42 4

COLMAPsoft		54.01 14	44.78 14	60.17 17	58.77 21	63.60 24	45.57 14	43.99 16	58.14 15

PLC	68.80 11	53.38 15	47.19 11	57.51 23	57.17 22	61.22 27	46.82 11	47.56 11	54.14 20
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
3Dnovator+	73.60 5	53.20 16	38.24 20	63.18 14	62.18 13	70.72 8	36.81 19	39.68 21	56.64 18

3Dnovator	73.91 4	52.81 17	38.17 21	62.58 15	60.95 16	69.63 11	35.98 23	40.36 20	57.16 17

ANet-0.75		52.71 18	55.63 2	50.76 32	44.38 34	59.75 28	53.05 1	58.21 2	48.17 27

ACMH	63.93 14	52.68 19	35.79 26	63.93 12	60.12 19	70.07 10	34.99 24	36.59 26	61.61 10
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
GSE		52.65 20	41.64 17	59.98 19	55.71 24	65.76 21	41.60 16	41.69 18	58.48 14

OpenMVS	70.45 8	51.01 21	37.34 22	60.13 18	56.52 23	67.91 14	33.72 25	40.96 19	55.96 19

CIDER		49.33 22	33.81 27	59.69 21	61.47 14	68.26 13	31.50 27	36.11 27	49.34 26
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
OpenMVS_ROB	61.12 15	48.48 23	31.63 28	59.71 20	59.03 20	67.15 16	31.08 28	32.18 30	52.95 23

LPCS		48.13 24	38.62 19	54.48 27	51.49 27	57.98 30	38.93 17	38.30 24	53.97 21

CPR_FA		47.15 25	40.41 18	51.65 31	47.19 31	53.80 36	36.70 21	44.12 15	53.96 22

COLMAP_ROB	57.96 17	45.89 26	37.03 24	51.79 30	49.43 30	56.42 35	36.78 20	37.28 25	49.53 25
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
LTVRE_ROB	59.60 16	43.60 27	37.31 23	47.80 33	45.94 33	49.52 41	36.19 22	38.43 23	47.93 28
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
unMVSv1		43.21 28	44.62 15	42.27 38	40.66 37	51.71 39	47.08 9	42.16 17	34.44 39

Snet		41.99 29	16.05 43	59.29 22	60.27 18	70.55 9	9.62 49	22.48 41	47.04 30

cscdsrtA		41.31 30	18.44 39	56.56 24	55.29 25	67.08 17	13.89 40	22.98 39	47.30 29

MVSNet_++		41.03 31	19.38 36	55.47 25	63.04 10	57.24 31	6.23 53	32.52 29	46.14 31

cscd		38.80 32	17.46 41	53.03 28	50.40 29	63.89 23	13.07 42	21.85 42	44.80 32

P-MVSNet		38.28 33	29.70 30	44.00 36	39.20 38	50.96 40	29.12 29	30.29 33	41.85 33

Pnet-new-		38.13 34	13.54 45	54.53 26	50.63 28	63.16 26	15.34 38	11.75 50	49.79 24

R-MVSNet		37.16 35	26.49 31	44.27 35	42.37 36	52.02 38	24.06 31	28.93 35	38.41 37

AttMVS		37.07 36	30.02 29	41.77 39	34.23 41	52.58 37	28.11 30	31.93 31	38.51 36

ANet		37.00 37	21.13 34	47.57 34	44.38 34	59.75 28	18.53 33	23.72 38	38.59 35

Pnet_fast		36.12 38	12.02 49	52.20 29	52.81 26	63.56 25	6.47 52	17.57 44	40.22 34

firsttry		32.39 39	25.47 32	36.99 40	32.51 42	46.62 42	21.58 32	29.36 34	31.86 41

MVSNet_plusplus		29.36 40	9.54 52	42.57 37	46.21 32	44.17 43	6.65 51	12.43 49	37.33 38

MVSNet		27.67 41	18.22 40	33.96 41	26.28 45	43.46 45	13.67 41	22.78 40	32.14 40

MVE	24.84 20	26.04 42	20.59 35	29.66 46	22.29 48	42.00 46	15.49 37	25.70 36	24.70 45
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
MVSCRF		24.97 43	13.50 46	32.61 44	26.31 44	43.85 44	12.02 44	14.98 46	27.67 42

example		24.50 44	16.56 42	29.80 45	27.47 43	56.53 34	17.69 34	15.43 45	5.40 54

Pnet-blend		23.49 45	9.16 53	33.04 42	38.23 39	35.75 49	3.77 55	14.54 47	25.15 43

Pnet-blend++		23.49 45	9.16 53	33.04 42	38.23 39	35.75 49	3.77 55	14.54 47	25.15 43

hgnet		22.90 47	13.03 47	29.49 47	25.13 46	56.72 32	16.70 35	9.35 53	6.62 52

DPSNet		22.90 47	13.03 47	29.49 47	25.13 46	56.72 32	16.70 35	9.35 53	6.62 52

cscdsrtB		21.94 49	19.24 37	23.74 51	21.33 49	34.30 51	14.54 39	23.93 37	15.57 48

F/T MVSNet+Gipuma		19.68 50	11.34 50	25.23 49	16.34 51	38.41 47	11.15 46	11.54 51	20.95 46

MVSNet + Gipuma		19.59 51	11.32 51	25.11 50	16.29 53	38.21 48	11.18 45	11.45 52	20.83 47

RMVSNet		15.61 52	22.67 33	10.90 54	13.04 55	14.34 55	12.27 43	33.07 28	5.30 55

Pnet-eth		14.99 53	18.95 38	12.35 53	19.54 50	4.23 58	7.59 50	30.32 32	13.27 49

PMVS	26.43 19	13.46 54	7.26 55	17.59 52	14.19 54	30.85 52	10.42 47	4.11 55	7.73 51
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
metmvs_fine		12.14 55	14.21 44	10.76 55	8.78 56	13.01 56	10.35 48	18.07 43	10.49 50

unMVSmet		7.50 56	3.60 56	10.10 56	6.98 57	18.02 54	3.83 54	3.37 56	5.28 56

CMPMVS	48.56 18	4.84 57	0.02 59	8.06 57	1.22 59	22.96 53	0.03 59	0.00 59	0.00 59
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
confMetMVS		4.71 58	2.07 57	6.46 58	4.51 58	11.02 57	1.90 57	2.24 57	3.85 57

FADENet		0.32 59	0.12 58	0.45 59	0.50 60	0.77 59	0.16 58	0.09 58	0.08 58

dnet		0.00 60	0.00 60	0.00 60	0.00 61	0.00 60	0.00 60	0.00 59	0.00 59

UnsupFinetunedMVSNet					16.34 51

Low-resolution many-view benchmark