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

DeepPCF-MVS	80.84 1	63.41 1	53.95 1	69.71 2	69.72 4	71.26 1	48.73 1	59.18 1	68.16 2

DeepC-MVS	79.81 2	62.37 2	52.77 2	68.77 3	68.65 5	70.30 3	46.95 2	58.59 3	67.37 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	79.65 3	62.24 3	50.96 4	69.75 1	69.91 3	71.08 2	46.04 3	55.88 4	68.27 1
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
3Dnovator+	77.84 4	59.53 4	47.18 8	67.76 4	70.63 1	67.68 4	43.34 7	51.01 10	64.98 7

3Dnovator	76.31 5	59.36 5	48.19 5	66.81 5	70.04 2	67.59 5	44.27 5	52.10 8	62.80 11

tm-dncc		58.94 6	48.02 7	66.22 6	66.30 8	65.64 8	41.30 12	54.73 5	66.71 5

TAPA-MVS	73.13 9	58.67 7	52.34 3	62.89 13	63.68 12	63.46 13	45.85 4	58.83 2	61.53 12
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
PCF-MVS	73.52 7	57.06 8	48.10 6	63.03 11	66.85 7	62.36 15	43.32 8	52.89 6	59.86 15
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMP	74.13 6	56.41 9	44.60 13	64.28 7	61.46 16	65.81 7	37.07 20	52.13 7	65.56 6
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
COLMAP(SR)		56.21 10	44.16 14	64.25 8	63.87 11	62.07 16	41.86 10	46.45 18	66.80 4

OpenMVS	72.83 10	56.18 11	45.66 10	63.19 10	64.67 10	65.13 10	39.60 15	51.71 9	59.76 16

COLMAP(base)		55.51 12	44.97 12	62.54 15	60.85 17	62.46 14	42.24 9	47.71 14	64.29 8

TAPA-MVS(SR)		55.07 13	43.33 16	62.90 12	67.41 6	57.81 25	39.94 14	46.71 17	63.49 9

ACMM	73.20 8	55.01 14	43.19 17	62.89 13	59.60 19	66.07 6	35.89 23	50.48 11	63.01 10
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
tmmvs		53.89 15	39.15 24	63.73 9	66.30 8	65.64 8	37.77 19	40.52 27	59.24 18

LTVRE_ROB	69.57 13	53.52 16	45.46 11	58.89 17	58.76 21	60.60 20	43.91 6	47.01 16	57.32 22
Andreas Kuhn, Heiko HirschmÃ¼ller, Daniel Scharstein, Helmut Mayer: A TV Prior for High-Quality Scalable Multi-View Stereo Reconstruction. International Journal of Computer Vision 2016
PLC	70.83 11	53.50 17	45.79 9	58.64 19	56.31 28	60.76 18	41.70 11	49.88 12	58.86 19
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
GSE		52.52 18	40.81 20	60.34 16	61.71 15	58.23 23	38.13 18	43.48 22	61.07 14

COLMAP_ROB	66.92 17	52.32 19	42.45 18	58.89 17	56.18 29	61.09 17	38.61 16	46.28 19	59.41 17
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
ACMH+	68.96 14	52.02 20	43.83 15	57.49 22	59.85 18	51.27 38	40.04 13	47.62 15	61.35 13

LPCS		50.60 21	40.43 21	57.38 23	58.37 23	56.00 29	38.14 17	42.71 24	57.77 20

HY-MVS	69.67 12	49.80 22	40.39 22	56.07 26	59.17 20	53.68 33	36.13 21	44.66 20	55.36 25

IB-MVS	68.01 15	49.19 23	42.00 19	53.98 31	58.50 22	51.98 35	35.85 24	48.16 13	51.46 31
Christian Sormann, Mattia Rossi, Andreas Kuhn and Friedrich Fraundorfer: IB-MVS: An Iterative Algorithm for Deep Multi-View Stereo based on Binary Decisions. BMVC 2021
OpenMVS_ROB	64.09 19	48.56 24	38.68 25	55.15 27	61.90 14	49.65 40	36.02 22	41.33 26	53.89 27

ACMH	67.68 16	47.97 25	38.24 27	54.45 30	56.88 25	50.08 39	34.49 26	41.99 25	56.40 23
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
CIDER		47.21 26	38.36 26	53.11 33	55.37 31	55.10 32	33.25 29	43.46 23	48.87 37
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
PVSNet_0	57.27 20	46.73 27	34.96 30	54.57 29	55.50 30	51.93 36	31.22 30	38.71 29	56.27 24

AttMVS		45.85 28	39.26 23	50.25 39	49.36 40	51.75 37	34.83 25	43.70 21	49.63 35

PVSNet	64.34 18	45.78 29	36.40 29	52.03 34	52.49 34	49.20 42	34.18 28	38.61 30	54.41 26

test_1126		45.72 30	28.57 36	57.15 24	62.33 13	59.06 21	24.22 35	32.91 38	50.08 33

CasMVSNet(SR_A)		45.13 31	25.99 40	57.89 20	56.32 27	64.81 11	19.76 43	32.23 39	52.54 30

PVSNet_LR		44.83 32	30.05 32	54.69 28	53.21 32	57.88 24	21.84 38	38.25 31	52.98 28

CasMVSNet(base)		44.49 33	24.93 41	57.53 21	56.38 26	64.76 12	18.64 45	31.23 41	51.43 32

P-MVSNet		44.46 34	37.09 28	49.37 41	49.27 41	49.30 41	34.35 27	39.83 28	49.54 36

BP-MVSNet		43.22 35	32.08 31	50.65 38	52.86 33	46.16 49	27.25 32	36.92 32	52.94 29
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
mvs_zhu_1030		42.98 36	27.14 37	53.53 32	52.24 35	58.28 22	20.89 39	33.39 36	50.07 34

Pnet-new-		42.61 37	22.17 46	56.24 25	58.21 24	53.03 34	24.08 36	20.25 57	57.48 21

A-TVSNet + Gipuma		42.12 38	29.09 35	50.80 36	48.28 42	56.53 27	29.14 31	29.04 42	47.59 38

test_1205		40.46 39	29.35 34	47.87 42	47.88 43	56.48 28	24.37 34	34.32 34	39.23 46

ANet		39.53 40	24.60 42	49.48 40	49.42 39	55.75 30	22.08 37	27.13 46	43.25 41

ANet-0.75		39.47 41	22.61 45	50.71 37	49.43 38	55.75 30	18.21 46	27.01 47	46.96 39

MVSNet		38.33 42	26.99 38	45.89 43	40.49 48	57.57 26	20.52 41	33.47 35	39.63 45

Pnet_fast		37.64 43	17.70 52	50.93 35	49.72 37	60.73 19	10.59 69	24.81 50	42.34 42

R-MVSNet		36.87 44	29.78 33	41.61 47	42.00 47	46.99 44	24.73 33	34.83 33	35.83 47

CasMVSNet(SR_B)		32.16 45	26.80 39	35.73 50	34.57 51	46.99 44	20.55 40	33.05 37	25.63 56

test_1124		31.65 46	13.28 68	43.90 44	44.21 46	47.60 43	14.80 50	11.75 75	39.88 44

Pnet-blend++		31.51 47	15.02 61	42.51 45	47.41 44	46.50 47	6.99 76	23.05 51	33.62 49

Pnet-blend		31.51 47	15.02 61	42.51 45	47.41 44	46.50 47	6.99 76	23.05 51	33.62 49

unsupervisedMVS_cas		30.64 49	20.24 47	37.57 49	35.93 50	42.74 52	15.36 48	25.12 49	34.04 48

MVSNet_plusplus		28.64 50	12.93 69	39.11 48	49.80 36	21.55 69	6.17 79	19.69 58	45.98 40

MVSCRF		28.32 51	18.18 50	35.09 52	32.16 54	44.37 51	14.66 51	21.69 54	28.73 52

MVSNet_++		25.72 52	17.38 54	31.27 53	33.47 52	18.28 74	2.87 84	31.90 40	42.06 43

MVS_test_1		25.14 53	18.49 48	29.57 54	37.50 49	27.99 59	9.50 72	27.48 45	23.21 63

vp_mvsnet		23.86 54	6.73 79	35.28 51	31.55 55	46.94 46	6.47 78	6.99 78	27.34 53

Snet		23.85 55	17.41 53	28.15 55	32.72 53	24.90 60	12.22 62	22.60 53	26.83 54

A1Net		23.74 56	23.95 43	23.61 59	21.43 62	19.62 71	19.93 42	27.96 43	29.78 51

CPR_FA		23.28 57	23.62 44	23.05 60	19.33 64	23.00 64	19.34 44	27.89 44	26.81 55

PMVS	37.38 22	21.09 58	11.49 76	27.48 56	24.09 57	44.44 50	15.98 47	7.01 77	13.92 74
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
test3		20.33 59	17.79 51	22.02 61	17.88 66	24.31 61	14.91 49	20.68 56	23.87 57

TVSNet		20.06 60	17.25 55	21.93 62	18.25 65	23.77 62	13.74 56	20.77 55	23.78 58

test_mvsss		19.92 61	11.93 75	25.25 57	24.18 56	32.22 54	7.42 75	16.44 67	19.34 68

CCVNet		19.37 62	12.14 73	24.19 58	20.50 63	28.61 58	11.03 68	13.25 71	23.48 59

QQQNet		18.88 63	15.19 60	21.34 63	17.26 68	23.30 63	13.79 55	16.60 64	23.48 59

SGNet		18.50 64	16.17 57	20.06 66	16.84 69	22.87 65	13.42 57	18.92 61	20.47 66

SVVNet		18.01 65	14.28 65	20.49 64	15.75 74	22.30 66	11.95 64	16.60 64	23.43 61

ternet		18.01 65	14.28 65	20.49 64	15.75 74	22.30 66	11.95 64	16.60 64	23.43 61

PSD-MVSNet		17.93 67	15.59 58	19.48 67	16.64 72	22.17 68	13.02 58	18.15 63	19.64 67

F/T MVSNet+Gipuma		17.31 68	14.39 63	19.26 69	16.83 70	19.67 70	14.03 53	14.75 68	21.27 64

MVSNet + Gipuma		16.91 69	14.12 67	18.77 72	16.39 73	19.13 73	13.98 54	14.27 69	20.78 65

hgnet		16.45 70	12.31 71	19.21 70	21.51 60	30.31 56	12.80 59	11.82 73	5.81 78

DPSNet		16.45 70	12.31 71	19.21 70	21.51 60	30.31 56	12.80 59	11.82 73	5.81 78

MVE	26.22 23	16.26 72	16.97 56	15.79 74	11.75 77	19.40 72	14.45 52	19.48 59	16.21 70
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
Pnet-eth		16.08 73	18.30 49	14.59 75	21.78 59	7.07 81	9.66 70	26.95 48	14.91 72

firsttry		15.88 74	12.57 70	18.08 73	17.80 67	17.14 75	11.54 66	13.61 70	19.31 69

example		15.69 75	10.14 77	19.39 68	23.73 58	30.32 55	11.10 67	9.17 76	4.13 82

unMVSv1		13.05 76	12.03 74	13.73 76	14.18 76	16.18 76	12.13 63	11.93 72	10.84 75

RMVSNet		11.09 77	15.49 59	8.17 81	9.16 78	10.05 78	12.57 61	18.40 62	5.29 81

metmvs_fine		10.03 78	14.31 64	7.18 82	6.75 82	6.98 82	9.52 71	19.10 60	7.81 76

test_1120		9.53 79	6.82 78	11.33 78	8.74 79	9.06 79	7.57 73	6.06 79	16.20 71

unMVSmet		7.42 80	5.82 80	8.48 80	7.60 80	11.29 77	5.84 80	5.81 80	6.56 77

CMPMVS	51.72 21	7.38 81	0.03 84	12.27 77	2.37 85	34.46 53	0.06 86	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
Cas-MVS_preliminary		7.24 82	3.51 81	9.72 79	7.03 81	7.46 80	4.87 82	2.16 82	14.65 73

confMetMVS		4.86 83	3.39 82	5.85 83	5.59 83	6.61 83	2.94 83	3.83 81	5.35 80

FADENet		0.17 84	0.15 83	0.18 85	0.31 86	0.16 85	0.19 85	0.10 83	0.08 84

dnet		0.00 85	0.00 85	0.00 86	0.00 87	0.00 86	0.00 87	0.00 84	0.00 85

test_MVS							7.47 74

test_robustmvs				3.18 84	5.44 84	2.71 84	4.94 81		1.39 83

UnsupFinetunedMVSNet					16.83 70

Low-resolution many-view benchmark