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

IB-MVS	68.87 2	56.55 1	40.64 1	67.16 1	66.47 1	71.01 1	41.57 1	39.72 10	63.99 1
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
BP-MVSNet		53.49 2	37.45 6	64.19 2	63.33 2	65.79 3	34.82 3	40.09 8	63.44 2
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
HY-MVS	67.03 5	50.55 3	32.69 14	62.45 3	60.70 3	68.35 2	27.36 20	38.02 14	58.31 3

ACMP	61.11 9	48.80 4	37.04 8	56.63 5	52.54 9	65.40 4	33.30 9	40.78 5	51.96 5
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
PVSNet_0	57.04 13	47.10 5	35.09 12	55.11 7	58.42 4	61.36 10	29.44 15	40.74 6	45.56 10

A1Net		46.45 6	40.21 2	50.60 16	49.68 13	58.73 19	35.92 2	44.51 2	43.40 16

COLMAP(SR)		46.33 7	32.87 13	55.31 6	54.46 8	60.59 12	34.13 5	31.61 21	50.88 6

PVSNet	62.49 8	46.24 8	35.67 10	53.28 9	56.15 5	59.33 17	31.27 12	40.07 9	44.36 13

ACMH+	54.58 15	45.04 9	30.90 20	54.47 8	50.64 11	62.39 7	31.18 13	30.62 23	50.37 7

TAPA-MVS(SR)		44.65 10	24.21 28	58.28 4	55.25 7	65.37 5	21.37 29	27.05 30	54.23 4

ACMM	58.35 12	43.94 11	31.72 17	52.08 11	47.41 20	60.88 11	28.78 16	34.67 17	47.96 8
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
DeepC-MVS	67.15 4	43.75 12	35.57 11	49.21 18	47.75 18	55.55 23	32.62 10	38.51 13	44.34 14
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ANet-0.75		42.47 13	39.16 3	44.68 28	37.48 43	55.22 24	34.57 4	43.75 3	41.34 22

TAPA-MVS	56.12 14	42.44 14	38.70 4	44.93 27	46.06 22	46.61 40	32.53 11	44.86 1	42.13 19
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
DeepPCF-MVS	69.37 1	42.29 15	36.10 9	46.42 24	45.29 25	52.98 28	33.61 8	38.59 12	41.00 23

DeepC-MVS_fast	67.50 3	42.28 16	31.82 16	49.25 17	47.89 17	56.00 22	30.48 14	33.15 19	43.85 15
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
3Dnovator	64.70 6	41.27 17	26.80 23	50.91 14	48.70 16	59.54 15	24.41 22	29.19 25	44.50 12

A-TVSNet + Gipuma		40.74 18	37.57 5	42.84 33	42.36 34	48.89 32	34.10 6	41.05 4	37.29 33

PVSNet_LR		40.04 19	21.22 38	52.58 10	51.48 10	61.59 8	14.86 44	27.58 29	44.67 11

3Dnovator+	62.71 7	39.72 20	25.96 26	48.89 19	46.28 21	59.77 14	24.04 24	27.88 27	40.63 24

PSD-MVSNet		39.60 21	31.54 18	44.98 26	47.54 19	47.09 38	24.19 23	38.88 11	40.31 26

OpenMVS	61.00 11	39.41 22	26.72 24	47.87 22	44.49 29	57.11 20	22.70 27	30.74 22	42.02 20

PCF-MVS	61.03 10	39.26 23	37.17 7	40.66 40	41.36 36	46.83 39	33.84 7	40.50 7	33.78 39
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMH	53.70 16	39.10 24	21.42 37	50.89 15	45.54 24	60.27 13	21.26 30	21.58 40	46.85 9
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
OpenMVS_ROB	53.19 17	38.75 25	23.68 29	48.80 20	49.97 12	54.98 26	22.06 28	25.29 33	41.46 21

tm-dncc		37.79 26	31.36 19	42.07 35	38.37 39	48.76 33	28.38 17	34.34 18	39.09 29

SGNet		37.06 27	27.49 22	43.43 31	45.14 26	46.08 41	19.93 33	35.05 16	39.08 30

test3		36.90 28	25.78 27	44.31 29	45.04 27	48.67 35	18.96 38	32.60 20	39.21 28

CasMVSNet(SR_A)		36.65 29	14.24 51	51.60 12	48.76 14	63.22 6	10.30 56	18.18 51	42.81 17

test_1205		36.57 30	22.35 33	46.06 25	45.75 23	56.91 21	14.94 43	29.76 24	35.50 36

Snet		36.17 31	13.16 55	51.51 13	55.63 6	61.48 9	7.38 67	18.94 48	37.42 32

CPR_FA		35.79 32	32.09 15	38.26 45	31.76 50	42.67 52	27.86 18	36.31 15	40.36 25

GSE		34.50 33	22.84 32	42.28 34	41.15 37	47.70 36	23.95 25	21.72 39	37.99 31

CIDER		34.17 34	23.17 31	41.51 38	41.48 35	52.26 30	19.57 35	26.78 31	30.77 44
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
COLMAP(base)		34.10 35	26.39 25	39.24 42	38.06 41	45.59 43	27.22 21	25.56 32	34.07 38

CasMVSNet(base)		33.65 36	13.31 54	47.21 23	42.69 32	59.53 16	9.57 59	17.05 53	39.41 27

Pnet-new-		32.96 37	9.93 64	48.32 21	43.98 30	58.83 18	12.02 50	7.84 70	42.15 18

LPCS		32.50 38	23.32 30	38.62 44	37.92 42	43.60 50	23.52 26	23.12 35	34.33 37

TVSNet		32.41 39	21.56 36	39.64 41	40.47 38	44.98 45	14.65 45	28.48 26	33.49 41

PLC	52.38 18	31.95 40	27.79 21	34.73 48	34.11 47	42.31 53	27.78 19	27.80 28	27.78 49
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
unsupervisedMVS_cas		31.40 41	16.86 44	41.09 39	36.27 45	50.35 31	11.46 51	22.26 37	36.65 34

test_1126		31.36 42	12.22 58	44.11 30	42.64 33	54.01 27	9.64 58	14.81 60	35.69 35

ANet		31.21 43	15.05 49	41.98 37	37.47 44	55.22 24	13.56 47	16.53 56	33.26 42

tmmvs		31.18 44	19.93 40	38.68 43	38.37 39	48.76 33	19.56 36	20.30 41	28.92 48

MVSNet_++		30.63 45	11.85 59	43.15 32	48.73 15	47.10 37	3.97 74	19.73 43	33.61 40

QQQNet		28.92 46	18.55 42	35.84 47	43.16 31	45.56 44	17.40 39	19.71 44	18.79 58

Pnet_fast		28.71 47	8.76 67	42.02 36	44.61 28	52.33 29	4.71 72	12.81 63	29.11 47

AttMVS		28.16 48	21.62 35	32.52 50	27.99 58	43.04 51	19.00 37	24.25 34	26.52 50

LTVRE_ROB	45.45 19	27.98 49	21.63 34	32.22 52	29.17 54	38.00 57	20.46 32	22.79 36	29.50 45
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
mvs_zhu_1030		27.69 50	13.38 53	37.23 46	33.13 48	46.06 42	10.34 55	16.42 57	32.49 43

P-MVSNet		26.12 51	15.32 48	33.33 49	29.35 53	41.51 54	15.57 42	15.06 59	29.13 46

COLMAP_ROB	43.60 20	26.07 52	19.94 39	30.17 55	27.45 59	38.78 56	19.60 34	20.28 42	24.27 53
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
R-MVSNet		25.05 53	16.41 45	30.81 54	28.43 57	40.90 55	14.21 46	18.62 50	23.10 54

firsttry		24.45 54	19.64 41	27.66 59	23.54 63	36.55 60	17.10 40	22.17 38	22.89 55

SVVNet		22.86 55	15.45 46	27.79 56	28.76 55	34.35 66	11.20 52	19.71 44	20.26 56

ternet		22.86 55	15.45 46	27.79 56	28.76 55	34.35 66	11.20 52	19.71 44	20.26 56

MVS_test_1		22.25 57	13.97 52	27.78 58	32.56 49	37.39 59	10.28 57	17.66 52	13.38 65

test_1124		21.33 58	4.94 75	32.25 51	26.87 62	44.42 47	6.12 70	3.77 76	25.47 51

MVSNet_plusplus		21.00 59	4.58 76	31.94 53	35.15 46	35.23 64	3.04 77	6.11 71	25.45 52

unMVSv1		20.10 60	17.39 43	21.90 66	18.36 65	35.25 63	20.59 31	14.19 61	12.10 67

MVSNet		19.51 61	12.81 56	23.97 64	18.27 66	35.02 65	8.92 64	16.71 54	18.63 60

CCVNet		19.28 62	10.64 63	25.04 60	20.47 64	35.85 61	7.84 66	13.44 62	18.79 58

test_mvsss		18.59 63	9.35 66	24.75 61	27.17 61	32.39 68	9.26 62	9.45 65	14.68 63

MVSCRF		17.10 64	9.37 65	22.26 65	16.98 67	35.69 62	8.18 65	10.56 64	14.12 64

Pnet-blend++		16.90 65	5.61 73	24.42 62	29.94 51	26.63 73	2.00 80	9.23 67	16.70 61

Pnet-blend		16.90 65	5.61 73	24.42 62	29.94 51	26.63 73	2.00 80	9.23 67	16.70 61

example		16.74 67	11.31 61	20.36 68	14.87 69	44.64 46	13.29 49	9.33 66	1.58 82

CasMVSNet(SR_B)		16.51 68	14.73 50	17.70 71	14.75 70	28.18 71	10.82 54	18.64 49	10.17 70

MVE	16.60 23	14.85 69	12.50 57	16.41 72	9.29 73	27.71 72	9.15 63	15.86 58	12.25 66
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
hgnet		14.36 70	6.98 68	19.28 69	11.25 71	44.33 48	9.49 60	4.46 74	2.27 77

DPSNet		14.36 70	6.98 68	19.28 69	11.25 71	44.33 48	9.49 60	4.46 74	2.27 77

vp_mvsnet		13.26 72	1.90 79	20.84 67	14.99 68	37.43 58	2.65 78	1.15 80	10.11 71

F/T MVSNet+Gipuma		12.37 73	6.37 71	16.37 73	8.99 74	29.20 69	7.11 69	5.63 72	10.92 68

MVSNet + Gipuma		12.18 74	6.27 72	16.12 74	8.79 76	28.79 70	7.19 68	5.36 73	10.77 69

Pnet-eth		8.19 75	11.46 60	6.00 78	8.77 77	1.65 84	3.88 75	19.04 47	7.59 72

PMVS	19.57 22	6.39 76	2.66 77	8.87 76	5.91 78	19.30 76	3.83 76	1.49 78	1.40 83
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
RMVSNet		6.32 77	10.95 62	3.23 83	4.05 79	3.98 82	5.23 71	16.67 55	1.67 81

metmvs_fine		5.54 78	6.47 70	4.92 81	3.27 81	7.31 80	4.59 73	8.35 69	4.18 76

CMPMVS	40.41 21	4.58 79	0.01 84	7.64 77	1.00 85	21.91 75	0.02 86	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
unMVSmet		3.77 80	1.46 80	5.32 79	3.25 82	10.48 78	1.71 83	1.21 79	2.22 79

test_1120		3.60 81	2.15 78	4.57 82	3.54 80	3.18 83	2.58 79	1.72 77	6.98 73

Cas-MVS_preliminary		3.51 82	1.08 81	5.13 80	3.15 83	7.63 79	1.77 82	0.40 82	4.62 75

confMetMVS		2.21 83	0.77 82	3.18 84	2.40 84	5.36 81	0.69 84	0.85 81	1.77 80

FADENet		0.08 84	0.02 83	0.11 85	0.13 86	0.18 85	0.03 85	0.02 83	0.02 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							13.56 47

test_robustmvs				15.01 75	27.45 59	11.04 77	16.58 41		6.55 74

UnsupFinetunedMVSNet					8.99 74

Low-resolution many-view benchmark