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	93.97 1	87.63 1	82.10 1	91.31 1	91.80 1	91.34 1	77.64 1	86.56 1	90.79 2

DeepC-MVS_fast	93.89 2	86.59 2	80.39 2	90.73 2	91.00 3	90.77 2	75.90 3	84.89 4	90.41 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
3Dnovator+	91.43 4	81.91 17	73.11 21	87.78 8	90.06 4	90.17 3	70.40 17	75.82 24	83.12 19

3Dnovator	91.36 5	82.19 16	74.09 18	87.59 11	89.83 5	90.10 4	71.54 14	76.63 21	82.84 22

DeepC-MVS	93.07 3	85.88 3	79.73 5	89.98 3	89.82 6	90.09 5	74.93 5	84.53 5	90.03 4
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
OpenMVS	89.19 12	80.38 21	71.40 23	86.37 16	87.42 12	89.29 6	67.70 22	75.10 25	82.40 24

PCF-MVS	89.48 11	84.51 6	76.96 11	89.54 4	91.30 2	88.76 7	73.35 9	80.56 14	88.57 5
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMP	89.59 10	82.31 13	75.00 16	87.18 12	84.49 24	88.57 8	68.43 20	81.58 12	88.49 7
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
TAPA-MVS	90.10 7	84.80 5	80.13 4	87.91 7	88.22 8	88.54 9	75.35 4	84.91 3	86.96 15
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
COLMAP(SR)		83.28 11	76.24 14	87.97 6	87.61 11	87.90 10	72.76 12	79.71 16	88.39 8

COLMAP(base)		83.59 8	77.50 10	87.65 10	86.74 15	87.68 11	73.30 10	81.70 11	88.52 6

ACMM	89.79 8	82.20 15	75.34 15	86.76 15	84.94 21	87.68 11	68.85 19	81.84 10	87.67 12
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PLC	91.00 6	83.54 9	78.38 6	86.98 14	85.38 19	87.33 13	73.45 7	83.31 6	88.23 10
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
GSE		83.39 10	77.82 9	87.10 13	85.70 18	87.27 14	73.41 8	82.23 9	88.34 9

tm-dncc		85.31 4	80.20 3	88.72 5	87.91 9	87.16 15	74.75 6	85.64 2	91.10 1

tmmvs		77.36 24	66.00 27	84.93 21	87.91 9	87.16 15	64.41 26	67.60 31	79.73 27

COLMAP_ROB	87.81 15	81.57 18	74.74 17	86.13 18	83.81 25	87.08 17	69.46 18	80.01 15	87.49 14
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
TAPA-MVS(SR)		83.88 7	78.08 8	87.74 9	88.99 7	86.49 18	73.17 11	82.99 7	87.76 11

HY-MVS	89.66 9	79.55 22	71.86 22	84.69 22	84.86 22	86.32 19	66.45 23	77.26 20	82.88 21

A-TVSNet + Gipuma		77.14 25	67.96 26	83.26 24	80.53 28	86.26 20	65.25 24	70.67 26	82.99 20

test_1205		70.75 32	62.35 33	76.35 35	78.70 32	85.03 21	55.63 34	69.08 27	65.34 45

IB-MVS	87.33 17	81.13 19	76.89 12	83.96 23	86.75 14	84.70 22	71.43 15	82.36 8	80.43 25
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
ACMH+	87.92 14	82.46 12	76.63 13	86.35 17	86.95 13	84.52 23	71.68 13	81.58 12	87.56 13

test_1126		71.43 31	57.57 38	80.67 28	85.16 20	84.45 24	55.16 35	59.98 42	72.40 34

CasMVSNet(SR_A)		63.08 41	45.33 49	74.92 36	74.36 36	83.73 25	37.04 55	53.62 50	66.66 43

CIDER		72.92 30	62.80 32	79.66 30	80.16 30	83.22 26	60.69 31	64.90 37	75.61 32
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
LPCS		78.54 23	73.62 20	81.82 26	79.59 31	83.19 27	71.25 16	75.99 23	82.69 23

Pnet_fast		59.94 43	35.26 73	76.39 34	74.15 37	82.99 28	25.02 74	45.51 61	72.03 35

LTVRE_ROB	88.41 13	82.23 14	78.11 7	84.97 20	86.24 16	82.83 29	76.93 2	79.30 17	85.84 17
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
ACMH	87.59 16	80.55 20	73.69 19	85.11 19	86.17 17	82.81 30	68.19 21	79.20 18	86.36 16
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
CasMVSNet(base)		61.88 42	43.72 52	73.99 38	74.03 38	81.95 31	34.99 60	52.45 52	65.99 44

PVSNet_LR		70.64 33	59.04 36	78.38 32	77.15 35	81.75 32	50.11 39	67.97 29	76.22 29

OpenMVS_ROB	81.14 20	73.28 29	63.11 31	80.06 29	82.70 26	81.62 33	61.11 28	65.12 36	75.87 30

PVSNet_0	82.17 19	74.27 27	64.50 29	80.78 27	81.77 27	80.64 34	61.11 28	67.89 30	79.92 26

ANet-0.75		67.51 36	58.66 37	73.41 39	70.12 42	79.37 35	50.72 37	66.61 33	70.74 37

ANet		63.85 39	52.65 40	71.32 41	70.12 42	79.37 35	48.53 40	56.77 46	64.48 46

Pnet-new-		63.42 40	42.13 57	77.61 33	77.71 34	79.30 37	41.06 44	43.20 67	75.82 31

mvs_zhu_1030		64.43 38	51.10 41	73.32 40	71.88 41	78.97 38	42.87 41	59.32 43	69.10 40

PVSNet	86.66 18	73.29 28	65.21 28	78.67 31	80.40 29	77.98 39	64.04 27	66.38 34	77.62 28

R-MVSNet		70.54 34	63.98 30	74.91 37	73.85 39	77.87 40	59.85 32	68.11 28	73.01 33

BP-MVSNet		76.76 26	68.58 25	82.22 25	84.51 23	77.66 41	60.98 30	76.18 22	84.48 18
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
AttMVS		65.89 37	60.14 34	69.73 43	64.68 49	75.04 42	58.80 33	61.47 40	69.47 39

vp_mvsnet		45.13 60	25.47 79	58.23 55	53.74 60	73.85 43	19.74 80	31.20 76	47.10 63

P-MVSNet		70.42 35	71.20 24	69.90 42	66.73 46	72.02 44	64.84 25	77.55 19	70.95 36

unsupervisedMVS_cas		57.32 45	44.73 51	65.71 45	63.76 50	71.21 45	36.03 57	53.44 51	62.15 49

test_1124		50.28 55	31.72 75	62.65 51	58.29 54	70.82 46	27.23 72	36.21 75	58.84 50

Pnet-blend++		53.68 51	37.04 66	64.76 46	68.91 44	70.02 47	24.13 75	49.96 57	55.36 55

Pnet-blend		53.68 51	37.04 66	64.76 46	68.91 44	70.02 47	24.13 75	49.96 57	55.36 55

test_mvsss		54.02 50	40.84 58	62.81 49	62.37 51	69.60 49	27.56 70	54.13 47	56.46 53

MVSCRF		51.99 53	37.64 63	61.55 52	58.88 52	68.35 50	34.99 60	40.28 70	57.43 52

PMVS	53.92 22	42.74 71	26.50 77	53.57 57	54.03 59	68.05 51	36.35 56	16.65 82	38.64 69
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
MVSNet		56.22 47	46.49 45	62.70 50	57.79 55	67.16 52	38.93 48	54.05 48	63.16 48

CasMVSNet(SR_B)		48.99 57	45.75 47	51.15 59	49.07 66	62.78 53	37.57 49	53.93 49	41.61 67

MVS_test_1		56.68 46	46.21 46	63.66 48	72.26 40	62.56 54	31.06 65	61.35 41	56.15 54

hgnet		44.61 65	38.48 61	48.69 60	57.33 56	60.08 55	37.17 53	39.80 72	28.65 78

DPSNet		44.61 65	38.48 61	48.69 60	57.33 56	60.08 55	37.17 53	39.80 72	28.65 78

example		40.49 74	30.54 76	47.11 64	58.75 53	58.88 57	29.97 66	31.12 77	23.71 82

CPR_FA		58.26 44	56.35 39	59.54 54	56.43 58	55.34 58	50.61 38	62.09 39	66.83 42

MVE	50.73 23	44.67 62	45.57 48	44.07 71	35.90 79	52.30 59	40.89 45	50.26 55	44.00 66
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
Snet		49.32 56	36.02 72	58.18 56	66.02 47	50.24 60	27.93 69	44.12 62	58.27 51

unMVSv1		45.02 61	43.12 54	46.29 65	47.83 67	50.14 61	42.67 42	43.57 63	40.92 68

MVSNet_plusplus		54.10 49	36.12 71	66.09 44	78.36 33	49.55 62	20.40 79	51.84 53	70.36 38

CCVNet		38.86 76	31.75 74	43.59 72	44.13 69	49.27 63	25.48 73	38.01 74	37.38 72

MVSNet_++		50.85 54	37.40 64	59.81 53	65.34 48	46.28 64	9.19 84	65.60 35	67.81 41

F/T MVSNet+Gipuma		43.59 70	37.31 65	47.78 62	50.12 63	45.64 65	34.25 63	40.37 69	47.60 62

MVSNet + Gipuma		44.11 69	38.76 60	47.67 63	50.05 65	45.26 66	34.93 62	42.60 68	47.71 61

test3		45.71 58	44.95 50	46.21 66	42.67 72	44.26 67	39.04 47	50.87 54	51.69 57

firsttry		41.11 72	36.23 68	44.36 70	43.82 70	43.77 68	32.28 64	40.18 71	45.51 65

PSD-MVSNet		44.35 68	43.10 55	45.19 68	42.64 73	43.61 69	37.46 51	48.74 60	49.32 60

SGNet		44.49 67	43.48 53	45.16 69	42.20 75	43.60 70	37.53 50	49.43 59	49.69 59

TVSNet		44.65 64	42.86 56	45.84 67	43.41 71	43.54 71	35.52 58	50.20 56	50.56 58

QQQNet		40.79 73	40.37 59	41.07 74	42.37 74	43.46 72	37.29 52	43.44 64	37.38 72

A1Net		54.85 48	59.98 35	51.43 58	47.46 68	42.97 73	52.72 36	67.24 32	63.86 47

unMVSmet		32.09 79	25.89 78	36.22 78	36.96 78	42.02 74	23.79 77	27.99 78	29.68 77

RMVSNet		44.67 62	50.67 42	40.67 75	50.79 62	41.21 75	42.32 43	59.02 44	30.02 76

SVVNet		37.41 77	36.23 68	38.19 76	37.28 76	38.77 76	29.02 67	43.44 64	38.53 70

ternet		37.41 77	36.23 68	38.19 76	37.28 76	38.77 76	29.02 67	43.44 64	38.53 70

CMPMVS	62.92 21	8.44 83	0.22 84	13.92 83	4.26 85	37.50 78	0.45 86	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
test_1120		27.94 80	22.22 81	31.76 80	20.92 84	37.22 79	18.54 83	25.90 80	37.14 74

Cas-MVS_preliminary		25.33 82	19.98 82	28.90 81	24.65 82	34.26 80	18.62 82	21.35 81	27.78 80

confMetMVS		26.62 81	23.72 80	28.55 82	28.79 81	33.00 81	21.33 78	26.10 79	23.86 81

metmvs_fine		39.41 75	48.46 44	33.38 79	35.53 80	31.67 82	39.28 46	57.64 45	32.95 75

Pnet-eth		45.62 59	49.15 43	43.27 73	53.31 61	29.96 83	35.03 59	63.26 38	46.53 64

test_robustmvs				13.65 84	22.17 83	12.51 84	19.49 81		6.28 83

FADENet		1.48 84	1.79 83	1.27 85	2.11 86	1.06 85	2.23 85	1.34 83	0.65 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							27.53 71

UnsupFinetunedMVSNet					50.12 63

Low-resolution many-view benchmark