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

DeepC-MVS_fast	68.24 3	46.61 3	34.94 4	54.40 1	55.37 3	57.55 1	31.02 4	38.86 4	50.28 1
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
COLMAP(SR)		41.20 10	29.58 11	48.95 8	48.70 11	48.18 14	28.26 8	30.89 19	49.98 2

DeepC-MVS	69.38 2	47.34 1	38.25 1	53.40 2	54.29 4	56.62 3	32.95 1	43.55 2	49.30 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMP	63.53 6	42.50 6	31.11 10	50.08 6	48.62 12	52.78 7	24.90 18	37.33 8	48.84 4
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
DeepPCF-MVS	69.58 1	46.90 2	37.52 3	53.15 3	53.58 5	57.05 2	32.83 2	42.22 3	48.83 5

3Dnovator+	66.72 4	44.84 5	33.14 7	52.63 4	56.23 2	53.78 4	29.20 6	37.07 10	47.89 6

TAPA-MVS(SR)		40.72 12	29.34 12	48.31 11	52.32 6	44.77 24	26.63 11	32.05 14	47.84 7

tm-dncc		42.17 8	31.89 9	49.03 7	49.63 8	50.53 11	26.59 12	37.19 9	46.92 8

3Dnovator	64.47 5	45.24 4	34.37 5	52.48 5	57.34 1	53.53 5	30.63 5	38.11 6	46.56 9

ACMM	61.98 7	40.80 11	29.05 13	48.64 9	46.87 14	53.22 6	23.47 22	34.63 11	45.82 10
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
Pnet-new-		32.65 33	16.09 43	43.68 17	46.15 19	39.98 31	18.51 31	13.67 55	44.91 11

ACMH+	57.40 11	36.48 17	28.86 14	41.57 24	43.89 22	37.61 38	26.39 14	31.33 17	43.20 12

OpenMVS	61.03 9	42.28 7	32.86 8	48.56 10	51.94 7	50.82 10	27.02 9	38.70 5	42.92 13

CasMVSNet(SR_A)		35.87 19	19.38 34	46.87 12	46.17 18	51.96 9	14.32 41	24.45 31	42.47 14

COLMAP(base)		37.35 15	28.09 20	43.52 18	42.34 24	46.68 18	26.55 13	29.63 22	41.53 15

TAPA-MVS	59.36 10	41.97 9	37.54 2	44.93 15	46.78 16	46.52 19	31.44 3	43.65 1	41.48 16
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
GSE		34.79 23	23.63 29	42.23 23	46.31 17	39.18 33	23.20 24	24.07 32	41.20 17

CasMVSNet(base)		35.32 21	18.44 38	46.56 13	46.00 20	52.64 8	13.32 42	23.57 35	41.04 18

PVSNet_LR		34.22 26	21.35 30	42.80 21	42.24 26	45.98 21	14.83 39	27.87 24	40.18 19

HY-MVS	56.14 13	35.94 18	28.51 17	40.89 25	43.99 21	38.59 36	25.07 16	31.96 15	40.09 20

mvs_zhu_1030		33.71 27	19.38 34	43.27 19	42.20 27	47.82 16	15.18 37	23.58 34	39.77 21

ACMH	55.70 15	32.91 32	23.69 28	39.06 31	41.22 28	36.76 42	21.84 27	25.54 29	39.18 22
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PVSNet_0	43.31 20	33.35 29	23.83 27	39.69 27	41.22 28	38.75 35	21.08 28	26.59 26	39.10 23

tmmvs		38.38 14	26.53 22	46.29 14	49.63 8	50.53 11	25.02 17	28.03 23	38.71 24

PCF-MVS	61.88 8	40.12 13	33.16 6	44.75 16	49.39 10	46.16 20	28.63 7	37.70 7	38.70 25
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
LTVRE_ROB	55.42 16	36.88 16	28.57 16	42.42 22	40.72 30	48.22 13	26.97 10	30.17 21	38.32 26
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
OpenMVS_ROB	52.78 18	35.39 20	27.98 21	40.33 26	46.81 15	36.10 44	24.79 19	31.17 18	38.08 27

PVSNet	50.76 19	32.44 35	24.74 26	37.57 37	38.43 35	36.66 43	23.03 25	26.46 27	37.62 28

LPCS		33.63 28	24.80 25	39.52 28	42.29 25	39.01 34	23.44 23	26.15 28	37.27 29

BP-MVSNet		29.92 37	20.53 32	36.17 39	37.89 36	33.68 49	17.55 32	23.52 36	36.95 30
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
test_1126		33.11 30	18.21 40	43.04 20	47.43 13	45.33 22	15.33 36	21.10 40	36.35 31

IB-MVS	56.42 12	34.86 22	28.22 19	39.29 30	43.59 23	38.19 37	23.59 21	32.85 13	36.09 32
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
ANet-0.75		29.06 39	13.44 46	39.47 29	39.32 33	44.06 26	10.78 47	16.11 49	35.02 33

AttMVS		34.35 25	28.45 18	38.28 34	40.03 31	40.06 30	23.67 20	33.23 12	34.75 34

PLC	56.13 14	34.63 24	28.70 15	38.59 33	36.78 38	44.65 25	26.03 15	31.36 16	34.33 35
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
COLMAP_ROB	52.97 17	33.06 31	25.38 24	38.18 35	35.81 40	45.04 23	22.92 26	27.84 25	33.69 36
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
ANet		30.02 36	16.73 42	38.88 32	39.32 33	44.06 26	15.61 35	17.85 43	33.26 37

P-MVSNet		29.35 38	20.98 31	34.93 40	34.55 44	36.99 40	20.06 30	21.91 39	33.25 38

MVSNet_plusplus		19.10 50	6.64 71	27.41 48	36.63 39	14.19 69	3.12 79	10.15 64	31.41 39

CIDER		32.55 34	25.75 23	37.08 38	39.65 32	40.58 29	20.91 29	30.59 20	31.00 40
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
test_1124		23.88 44	8.92 64	33.86 42	35.80 41	36.79 41	10.99 46	6.84 72	28.99 41

Pnet_fast		27.80 41	12.25 48	38.17 36	37.73 37	47.91 15	7.47 62	17.03 44	28.88 42

A-TVSNet + Gipuma		26.68 43	16.93 41	33.18 44	31.29 46	39.82 32	17.28 33	16.57 47	28.43 43

MVSNet_++		16.58 52	10.34 55	20.75 53	22.25 52	11.60 72	1.64 83	19.05 41	28.39 44

test_1205		28.37 40	19.29 36	34.43 41	34.50 45	41.99 28	15.81 34	22.78 37	26.79 45

MVSNet		27.45 42	18.54 37	33.39 43	29.32 47	47.13 17	13.18 43	23.89 33	23.72 46

unsupervisedMVS_cas		21.41 49	13.42 47	26.73 49	25.53 49	31.05 52	10.34 48	16.50 48	23.60 47

Pnet-blend++		21.89 47	9.21 62	30.34 45	34.58 42	34.44 45	3.76 76	14.67 52	22.00 48

Pnet-blend		21.89 47	9.21 62	30.34 45	34.58 42	34.44 45	3.76 76	14.67 52	22.00 48

R-MVSNet		23.84 45	18.30 39	27.54 47	27.82 48	33.79 48	14.57 40	22.03 38	21.00 50

A1Net		14.51 56	14.40 45	14.59 58	13.30 59	12.51 70	11.95 44	16.85 45	17.96 51

CasMVSNet(SR_B)		23.80 46	19.79 33	26.47 50	24.88 50	37.51 39	14.94 38	24.65 30	17.01 52

CCVNet		13.25 58	7.81 67	16.88 56	13.52 57	20.14 55	7.96 60	7.67 69	16.97 53

QQQNet		12.59 61	9.43 58	14.70 57	10.88 66	16.24 63	8.64 54	10.22 61	16.97 53

Snet		16.01 53	11.97 50	18.70 54	22.09 53	17.12 59	8.46 55	15.48 51	16.90 55

SVVNet		12.22 62	9.22 60	14.22 63	10.32 70	15.62 65	8.22 58	10.22 61	16.70 56

ternet		12.22 62	9.22 60	14.22 63	10.32 70	15.62 65	8.22 58	10.22 61	16.70 56

CPR_FA		14.43 57	14.90 44	14.12 65	11.38 62	14.82 68	11.91 45	17.89 42	16.15 58

vp_mvsnet		15.26 54	2.83 79	23.55 51	21.00 54	34.21 47	3.45 78	2.22 79	15.43 59

test3		13.22 59	11.25 51	14.54 59	11.46 61	17.02 60	9.59 50	12.91 57	15.13 60

TVSNet		13.02 60	10.86 53	14.46 61	11.60 60	16.67 62	8.79 53	12.93 56	15.10 61

MVSCRF		18.49 51	12.01 49	22.81 52	20.51 55	32.84 50	9.61 49	14.41 54	15.08 62

MVS_test_1		15.08 55	10.97 52	17.82 55	23.84 51	16.69 61	5.32 70	16.61 46	12.93 63

SGNet		11.73 64	9.85 56	12.98 66	10.55 68	15.85 64	8.23 57	11.47 58	12.53 64

PSD-MVSNet		11.22 65	9.34 59	12.48 67	10.37 69	15.20 67	7.81 61	10.86 60	11.87 65

firsttry		9.57 69	7.40 69	11.02 72	11.05 63	10.54 75	6.97 65	7.82 68	11.49 66

F/T MVSNet+Gipuma		9.78 68	8.13 65	10.88 73	9.23 73	11.92 71	8.81 51	7.45 70	11.49 66

test_1120		6.40 78	4.10 78	7.94 76	6.69 77	5.86 78	4.95 72	3.26 77	11.26 68

MVSNet + Gipuma		9.47 71	7.91 66	10.52 74	8.91 75	11.46 73	8.80 52	7.01 71	11.18 69

test_mvsss		11.00 66	6.00 75	14.32 62	13.56 56	19.16 58	4.03 75	7.98 67	10.25 70

MVE	17.77 23	9.03 74	9.66 57	8.62 75	6.03 78	11.40 74	8.30 56	11.02 59	8.42 71
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
Pnet-eth		8.45 75	10.49 54	7.10 78	9.98 72	2.92 83	5.14 71	15.84 50	8.40 72

Cas-MVS_preliminary		4.12 81	1.79 81	5.67 79	4.49 79	4.19 79	2.85 80	0.74 82	8.32 73

unMVSv1		6.85 76	6.06 74	7.37 77	7.56 76	9.33 76	6.35 68	5.76 75	5.22 74

metmvs_fine		4.94 79	7.02 70	3.56 81	3.18 82	3.86 80	4.62 73	9.43 66	3.63 75

unMVSmet		3.60 82	2.44 80	4.38 80	3.82 81	6.35 77	2.69 81	2.18 80	2.97 76

PMVS	28.69 22	10.60 67	4.70 77	14.53 60	10.82 67	30.05 53	6.66 67	2.74 78	2.71 77
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
confMetMVS		2.34 83	1.33 82	3.02 83	3.03 83	3.47 81	1.14 84	1.52 81	2.56 78

hgnet		9.31 72	6.52 72	11.17 70	10.91 64	20.13 56	7.15 63	5.89 73	2.47 79

DPSNet		9.31 72	6.52 72	11.17 70	10.91 64	20.13 56	7.15 63	5.89 73	2.47 79

RMVSNet		4.93 80	7.73 68	3.06 82	3.93 80	3.22 82	5.95 69	9.52 65	2.04 81

example		9.54 70	5.97 76	11.92 68	13.40 58	20.72 54	6.89 66	5.05 76	1.63 82

test_robustmvs				1.65 84	2.89 84	1.35 84	2.69 81		0.72 83

FADENet		0.06 84	0.04 83	0.07 85	0.12 86	0.08 85	0.05 85	0.03 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

CMPMVS	42.80 21	6.77 77	0.01 84	11.27 69	1.93 85	31.86 51	0.03 86	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
test_MVS							4.13 74

UnsupFinetunedMVSNet					9.23 73

Low-resolution many-view benchmark