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

LTVRE_ROB	75.46 1	54.46 1	42.09 7	62.70 1	67.45 8	65.97 3	39.56 5	44.63 10	54.68 8
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
3Dnovator+	73.19 2	54.26 2	46.24 4	59.61 8	71.65 3	48.87 13	37.19 7	55.30 1	58.31 5

DeepPCF-MVS	71.07 5	53.29 3	39.34 11	62.59 2	65.59 10	61.81 4	32.09 12	46.60 8	60.36 1

3Dnovator	65.95 11	52.63 4	47.99 1	55.72 11	69.72 7	48.62 16	41.11 3	54.88 2	48.82 14

DeepC-MVS_fast	69.89 7	52.44 5	39.25 12	61.23 4	65.60 9	59.17 5	31.57 15	46.93 7	58.91 2
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
tmmvs		52.24 6	39.98 10	60.41 6	70.28 4	52.42 10	34.74 9	45.23 9	58.53 4

DeepC-MVS	72.44 4	51.91 7	41.70 9	58.72 10	62.91 12	57.73 7	33.27 11	50.12 6	55.51 6
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
CasMVSNet(SR_B)		49.60 8	30.23 26	62.52 3	79.42 1	56.08 8	24.08 29	36.37 23	52.07 10

tm-dncc		49.39 9	32.78 23	60.45 5	70.28 4	52.42 10	25.02 25	40.54 16	58.66 3

MVSNet		49.19 10	33.57 22	59.60 9	74.12 2	72.06 1	25.24 24	41.90 13	32.63 40

AttMVS		48.45 11	42.09 7	52.69 14	70.22 6	37.46 29	31.38 16	52.80 4	50.38 12

OpenMVS	62.51 15	47.53 12	42.80 6	50.68 16	62.38 13	45.79 22	33.37 10	52.22 5	43.87 21

COLMAP_ROB	72.78 3	46.46 13	36.00 16	53.43 12	51.48 20	53.72 9	27.60 21	44.40 11	55.08 7
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
mvs_zhu_1030		45.91 14	35.22 17	53.03 13	58.12 16	49.72 12	28.56 18	41.87 14	51.26 11

PMVS	70.70 6	44.31 15	21.09 42	59.79 7	64.24 11	67.85 2	25.32 23	16.87 51	47.28 16
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
test_1120		44.17 16	46.51 3	42.61 27	61.86 14	36.78 30	61.21 1	31.80 33	29.19 47

Pnet-new-		44.14 17	46.92 2	42.29 28	48.54 23	30.28 44	40.25 4	53.58 3	48.06 15

PCF-MVS	63.80 13	42.44 18	30.04 28	50.71 15	61.29 15	45.52 23	24.81 26	35.27 28	45.31 18
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
test_1124		42.03 19	45.75 5	39.55 33	53.61 17	31.40 42	54.10 2	37.41 20	33.63 37

COLMAP(base)		41.98 20	30.58 24	49.58 17	47.71 25	47.83 17	25.91 22	35.25 29	53.20 9

TAPA-MVS	65.27 12	41.55 21	36.46 15	44.94 21	47.53 26	46.42 21	30.41 17	42.51 12	40.86 28
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
test_1126		40.73 22	37.06 14	43.17 26	53.43 18	39.06 28	37.45 6	36.68 22	37.03 32

TAPA-MVS(SR)		40.23 23	37.32 13	42.17 29	49.69 22	34.04 35	35.33 8	39.31 18	42.79 24

CasMVSNet(base)		39.99 24	30.03 29	46.63 18	49.88 21	47.19 20	21.89 37	38.17 19	42.82 23

PLC	62.01 16	38.52 25	30.23 26	44.04 23	39.89 37	47.28 18	24.50 27	35.97 24	44.94 20
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMM	69.25 9	38.38 26	27.21 32	45.83 19	46.34 27	47.28 18	19.82 39	34.60 30	43.86 22
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
CasMVSNet(SR_A)		38.17 27	30.38 25	43.36 25	43.84 31	44.10 26	23.45 30	37.31 21	42.14 25

ACMP	69.50 8	37.98 28	27.15 34	45.19 20	45.25 28	44.25 24	19.88 38	34.43 31	46.08 17
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
Pnet-blend++		37.88 29	33.76 20	40.62 31	40.91 35	48.74 14	31.73 13	35.78 25	32.22 42

Pnet-blend		37.88 29	33.76 20	40.62 31	40.91 35	48.74 14	31.73 13	35.78 25	32.22 42

COLMAP(SR)		37.49 31	27.16 33	44.38 22	44.04 29	39.98 27	24.12 28	30.20 36	49.12 13

P-MVSNet		36.49 32	34.17 19	38.03 34	42.00 32	33.36 37	28.21 20	40.14 17	38.72 30

GSE		36.13 33	24.74 38	43.73 24	52.95 19	33.24 39	22.50 34	26.98 43	45.01 19

LPCS		35.46 34	26.73 36	41.28 30	47.80 24	35.29 34	23.36 31	30.10 37	40.76 29

OpenMVS_ROB	54.93 17	35.00 35	34.45 18	35.37 38	44.02 30	26.87 49	28.29 19	40.62 15	35.22 34

CIDER		32.10 36	29.05 30	34.13 41	37.98 40	33.17 40	22.44 36	35.66 27	31.24 44
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
ACMH+	66.64 10	31.68 37	27.48 31	34.49 40	38.73 38	26.92 47	22.87 33	32.08 32	37.82 31

PVSNet_LR		30.39 38	21.48 41	36.33 36	35.81 42	36.68 31	14.80 47	28.17 39	36.51 33

ACMH	63.62 14	30.30 39	26.87 35	32.59 43	37.65 41	26.45 51	22.46 35	31.28 34	33.67 36
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
ANet		29.81 40	18.89 45	37.10 35	41.36 33	36.66 32	18.37 40	19.40 48	33.27 38

Pnet_fast		29.80 41	21.76 40	35.17 39	32.69 45	44.17 25	18.13 42	25.38 44	28.65 48

HY-MVS	49.31 19	28.53 42	25.35 37	30.64 45	34.50 44	26.89 48	23.13 32	27.56 41	30.55 45

vp_mvsnet		27.00 43	17.92 47	33.05 42	35.04 43	31.50 41	4.95 72	30.89 35	32.60 41

PVSNet_0	36.71 22	26.11 44	18.07 46	31.47 44	31.85 47	28.31 46	16.41 45	19.73 47	34.24 35

IB-MVS	49.67 18	25.63 45	22.23 39	27.89 49	32.43 46	26.12 52	16.47 44	28.00 40	25.13 50
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
PVSNet	43.83 21	25.27 46	18.99 44	29.47 47	29.22 49	26.53 50	18.23 41	19.75 46	32.66 39

ANet-0.75		24.93 47	8.13 63	36.13 37	41.36 33	36.65 33	6.38 65	9.87 64	30.38 46

MVSNet_plusplus		24.27 48	16.59 50	29.38 48	38.24 39	8.88 71	3.21 79	29.98 38	41.02 27

test_1205		23.54 49	17.62 48	27.49 50	27.69 50	33.27 38	16.80 43	18.44 49	21.51 53

R-MVSNet		23.44 50	20.97 43	25.09 53	27.25 51	28.79 45	14.95 46	26.99 42	19.25 54

MVSCRF		21.37 51	17.18 49	24.16 54	25.89 53	30.42 43	11.65 49	22.71 45	16.19 56

BP-MVSNet		20.82 52	14.19 51	25.24 52	27.04 52	22.63 53	11.73 48	16.64 52	26.07 49
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
A-TVSNet + Gipuma		20.66 53	10.98 55	27.12 51	24.81 54	33.59 36	11.57 50	10.38 62	22.97 52

CMPMVS	48.73 20	17.96 54	0.21 83	29.80 46	31.01 48	58.39 6	0.41 85	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
unsupervisedMVS_cas		16.42 55	11.26 54	19.85 55	19.70 55	22.45 54	9.42 54	13.10 55	17.41 55

Cas-MVS_preliminary		13.10 56	6.45 74	17.53 56	7.84 68	2.89 80	7.20 59	5.71 78	41.86 26

MVSNet_++		13.01 57	9.72 58	15.20 57	14.41 57	6.62 76	1.03 84	18.42 50	24.57 51

MVS_test_1		12.27 58	9.64 60	14.02 58	18.81 56	10.74 61	3.59 78	15.68 53	12.52 61

Snet		11.53 59	11.50 52	11.55 60	13.79 58	9.94 66	9.91 53	13.09 56	10.91 65

Pnet-eth		11.01 60	10.58 57	11.30 61	11.57 60	12.95 60	7.59 57	13.56 54	9.39 68

CCVNet		10.60 61	6.72 71	13.19 59	10.10 63	14.00 55	8.08 55	5.36 79	15.47 57

F/T MVSNet+Gipuma		10.33 62	11.29 53	9.69 66	9.48 64	7.49 72	11.57 50	11.01 59	12.11 62

MVSNet + Gipuma		9.86 63	10.74 56	9.27 67	9.03 66	7.15 74	11.33 52	10.16 63	11.62 63

CPR_FA		9.09 64	9.72 58	8.67 73	6.93 71	8.97 70	7.57 58	11.87 57	10.10 66

QQQNet		8.84 65	6.32 75	10.52 62	6.22 76	9.88 67	5.75 68	6.90 71	15.47 57

SVVNet		8.80 66	6.70 72	10.20 63	6.29 74	10.11 64	6.49 62	6.90 71	14.21 59

ternet		8.80 66	6.70 72	10.20 63	6.29 74	10.11 64	6.49 62	6.90 71	14.21 59

A1Net		8.71 68	8.78 61	8.67 73	7.67 69	7.00 75	7.17 60	10.39 61	11.32 64

TVSNet		8.28 69	7.32 66	8.92 69	6.77 72	10.23 63	6.28 67	8.36 68	9.75 67

test3		8.15 70	7.23 67	8.75 72	6.57 73	10.32 62	6.42 64	8.05 69	9.38 69

hgnet		8.15 70	7.20 68	8.77 70	10.59 61	13.03 58	5.73 69	8.68 65	2.70 79

DPSNet		8.15 70	7.20 68	8.77 70	10.59 61	13.03 58	5.73 69	8.68 65	2.70 79

test_mvsss		8.00 73	4.74 79	10.17 65	9.03 66	13.60 56	2.57 81	6.90 71	7.88 70

example		7.10 74	4.06 81	9.13 68	12.19 59	13.49 57	4.65 74	3.46 82	1.70 82

SGNet		7.01 75	6.02 76	7.67 75	5.97 77	9.57 68	5.19 71	6.86 75	7.46 72

MVE	27.91 23	6.82 76	8.02 64	6.02 78	4.47 81	7.18 73	7.60 56	8.44 67	6.42 74
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
PSD-MVSNet		6.56 77	5.49 77	7.28 76	5.82 78	9.06 69	4.66 73	6.31 77	6.96 73

unMVSmet		6.24 78	8.77 62	4.56 79	4.65 80	4.56 79	6.35 66	11.19 58	4.46 76

firsttry		6.03 79	4.56 80	7.01 77	7.22 70	6.16 77	4.38 76	4.75 80	7.67 71

metmvs_fine		4.88 80	7.73 65	2.98 83	3.10 84	2.62 82	4.65 74	10.82 60	3.22 78

RMVSNet		4.54 81	6.78 70	3.04 82	3.81 83	2.71 81	6.89 61	6.66 76	2.61 81

confMetMVS		4.28 82	5.07 78	3.76 81	4.09 82	2.57 83	3.21 79	6.93 70	4.62 75

unMVSv1		4.17 83	3.68 82	4.49 80	4.76 79	5.38 78	3.75 77	3.62 81	3.33 77

FADENet		0.08 84	0.09 84	0.07 85	0.12 86	0.05 85	0.12 86	0.06 83	0.04 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							2.44 82

test_robustmvs				0.87 84	1.53 85	0.72 84	1.46 83		0.38 83

UnsupFinetunedMVSNet					9.48 64

Low-resolution many-view benchmark