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	77.80 4	67.29 1	53.24 6	76.65 1	77.55 1	78.10 1	53.88 1	52.59 9	74.31 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		64.34 2	50.31 9	73.69 2	74.99 2	71.90 11	46.42 12	54.19 8	74.18 2
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
A1Net		62.59 3	54.91 3	67.70 9	68.95 6	72.10 9	52.15 3	57.67 3	62.06 13

ACMP	71.68 10	62.33 4	51.03 8	69.86 5	66.41 9	74.93 2	47.36 8	54.70 7	68.24 3
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
A-TVSNet + Gipuma		61.02 5	56.34 1	64.14 15	64.32 12	66.91 23	52.09 4	60.60 1	61.19 16

COLMAP(SR)		60.84 6	47.20 12	69.93 4	69.97 5	72.99 6	47.93 7	46.47 15	66.85 4

DeepPCF-MVS	81.17 1	60.64 7	53.51 5	65.40 14	64.27 13	69.40 18	51.28 6	55.74 6	62.53 12

ACMH+	65.35 16	59.83 8	46.18 15	68.93 7	65.70 10	74.24 4	46.34 13	46.01 18	66.83 5

DeepC-MVS	77.85 3	59.42 9	49.21 10	66.23 13	64.01 14	70.42 15	46.20 15	52.23 10	64.26 9
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	79.48 2	59.14 10	48.44 11	66.27 11	64.81 11	70.57 13	46.73 11	50.15 11	63.44 10
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
PVSNet_0	68.08 15	59.14 10	42.70 19	70.10 3	74.33 3	74.52 3	38.12 21	47.28 14	61.46 15

TAPA-MVS	70.22 12	58.89 12	51.21 7	64.01 16	62.59 16	66.45 25	46.31 14	56.11 5	62.99 11
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
PCF-MVS	73.15 9	58.42 13	54.11 4	61.29 20	61.25 19	64.94 27	51.80 5	56.43 4	57.66 20
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMM	69.62 13	58.27 14	46.31 14	66.24 12	60.67 21	72.46 7	42.63 17	50.00 12	65.60 7
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
HY-MVS	76.49 5	57.81 15	40.66 21	69.25 6	68.59 7	73.61 5	35.90 27	45.41 19	65.54 8

PVSNet	73.49 8	57.76 16	43.28 18	67.41 10	70.29 4	72.10 9	40.29 19	46.26 17	59.84 17

TAPA-MVS(SR)		55.57 17	35.84 30	68.73 8	67.52 8	72.24 8	33.19 30	38.50 29	66.42 6

COLMAP(base)		54.01 18	44.78 16	60.17 21	58.77 25	63.60 29	45.57 16	43.99 21	58.14 19

PLC	68.80 14	53.38 19	47.19 13	57.51 27	57.17 26	61.22 33	46.82 10	47.56 13	54.14 24
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
3Dnovator+	73.60 7	53.20 20	38.24 25	63.18 18	62.18 17	70.72 12	36.81 22	39.68 27	56.64 22

3Dnovator	73.91 6	52.81 21	38.17 26	62.58 19	60.95 20	69.63 17	35.98 26	40.36 26	57.16 21

ANet-0.75		52.71 22	55.63 2	50.76 39	44.38 45	59.75 34	53.05 2	58.21 2	48.17 32

ACMH	63.93 17	52.68 23	35.79 31	63.93 17	60.12 23	70.07 16	34.99 28	36.59 34	61.61 14
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
GSE		52.65 24	41.64 20	59.98 23	55.71 30	65.76 26	41.60 18	41.69 24	58.48 18

OpenMVS	70.45 11	51.01 25	37.34 27	60.13 22	56.52 28	67.91 20	33.72 29	40.96 25	55.96 23

CIDER		49.33 26	33.81 33	59.69 25	61.47 18	68.26 19	31.50 32	36.11 35	49.34 30
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
OpenMVS_ROB	61.12 18	48.48 27	31.63 34	59.71 24	59.03 24	67.15 21	31.08 33	32.18 39	52.95 27

LPCS		48.13 28	38.62 24	54.48 32	51.49 37	57.98 36	38.93 20	38.30 32	53.97 25

PSD-MVSNet		47.65 29	39.40 23	53.15 34	56.48 29	53.80 43	32.51 31	46.29 16	49.16 31

CPR_FA		47.15 30	40.41 22	51.65 38	47.19 42	53.80 43	36.70 24	44.12 20	53.96 26

COLMAP_ROB	57.96 20	45.89 31	37.03 29	51.79 37	49.43 40	56.42 42	36.78 23	37.28 33	49.53 29
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
test_1205		45.54 32	30.23 36	55.74 29	56.56 27	66.59 24	21.99 40	38.47 30	44.08 41

SGNet		44.07 33	34.19 32	50.65 40	53.93 33	51.56 50	26.54 36	41.85 23	46.47 36

LTVRE_ROB	59.60 19	43.60 34	37.31 28	47.80 43	45.94 44	49.52 54	36.19 25	38.43 31	47.93 33
Andreas Kuhn, Heiko Hirschmüller, Daniel Scharstein, Helmut Mayer: A TV Prior for High-Quality Scalable Multi-View Stereo Reconstruction. International Journal of Computer Vision 2016
unMVSv1		43.21 35	44.62 17	42.27 50	40.66 49	51.71 49	47.08 9	42.16 22	34.44 51

test3		42.75 36	31.34 35	50.36 41	52.00 35	53.28 45	23.88 38	38.80 28	45.79 38

Snet		41.99 37	16.05 58	59.29 26	60.27 22	70.55 14	9.62 65	22.48 56	47.04 35

CasMVSNet(SR_A)		41.31 38	18.44 54	56.56 28	55.29 32	67.08 22	13.89 55	22.98 54	47.30 34

MVSNet_++		41.03 39	19.38 50	55.47 30	63.04 15	57.24 37	6.23 70	32.52 38	46.14 37

test_1126		40.23 40	19.47 49	54.08 33	55.58 31	62.13 32	15.66 47	23.27 52	44.52 40

CasMVSNet(base)		38.80 41	17.46 56	53.03 35	50.40 39	63.89 28	13.07 57	21.85 57	44.80 39

TVSNet		38.58 42	26.99 39	46.30 45	49.23 41	49.73 53	19.13 42	34.85 36	39.94 45

P-MVSNet		38.28 43	29.70 38	44.00 47	39.20 51	50.96 51	29.12 34	30.29 42	41.85 43

Pnet-new-		38.13 44	13.54 61	54.53 31	50.63 38	63.16 31	15.34 50	11.75 66	49.79 28

unsupervisedMVS_cas		37.45 45	21.88 44	47.84 42	43.96 47	56.86 38	15.49 48	28.27 45	42.70 42

R-MVSNet		37.16 46	26.49 40	44.27 46	42.37 48	52.02 47	24.06 37	28.93 44	38.41 49

AttMVS		37.07 47	30.02 37	41.77 51	34.23 56	52.58 46	28.11 35	31.93 40	38.51 48

ANet		37.00 48	21.13 45	47.57 44	44.38 45	59.75 34	18.53 43	23.72 51	38.59 47

Pnet_fast		36.12 49	12.02 65	52.20 36	52.81 34	63.56 30	6.47 69	17.57 60	40.22 44

QQQNet		34.46 50	23.92 42	41.49 52	51.88 36	50.62 52	22.82 39	25.03 47	21.97 61

mvs_zhu_1030		33.65 51	18.58 53	43.70 48	39.88 50	51.94 48	14.13 52	23.03 53	39.29 46

firsttry		32.39 52	25.47 41	36.99 54	32.51 57	46.62 56	21.58 41	29.36 43	31.86 53

MVSNet_plusplus		29.36 53	9.54 68	42.57 49	46.21 43	44.17 57	6.65 68	12.43 65	37.33 50

MVSNet		27.67 54	18.22 55	33.96 55	26.28 61	43.46 59	13.67 56	22.78 55	32.14 52

ternet		27.23 55	19.50 47	32.39 59	35.04 54	38.67 62	13.98 53	25.03 47	23.46 59

SVVNet		27.23 55	19.50 47	32.39 59	35.04 54	38.67 62	13.98 53	25.03 47	23.46 59

MVE	24.84 23	26.04 57	20.59 46	29.66 63	22.29 65	42.00 60	15.49 48	25.70 46	24.70 58
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
test_1124		25.42 58	7.36 71	37.46 53	31.97 58	49.24 55	8.21 66	6.51 71	31.19 54

MVSCRF		24.97 59	13.50 62	32.61 58	26.31 60	43.85 58	12.02 59	14.98 62	27.67 55

example		24.50 60	16.56 57	29.80 62	27.47 59	56.53 41	17.69 44	15.43 61	5.40 73

CCVNet		23.74 61	14.42 59	29.95 61	25.90 62	41.97 61	9.71 64	19.13 58	21.97 61

Pnet-blend		23.49 62	9.16 69	33.04 56	38.23 52	35.75 66	3.77 73	14.54 63	25.15 56

Pnet-blend++		23.49 62	9.16 69	33.04 56	38.23 52	35.75 66	3.77 73	14.54 63	25.15 56

DPSNet		22.90 64	13.03 63	29.49 64	25.13 63	56.72 39	16.70 45	9.35 69	6.62 71

hgnet		22.90 64	13.03 63	29.49 64	25.13 63	56.72 39	16.70 45	9.35 69	6.62 71

CasMVSNet(SR_B)		21.94 66	19.24 51	23.74 68	21.33 66	34.30 68	14.54 51	23.93 50	15.57 65

F/T MVSNet+Gipuma		19.68 67	11.34 66	25.23 66	16.34 68	38.41 64	11.15 61	11.54 67	20.95 63

MVSNet + Gipuma		19.59 68	11.32 67	25.11 67	16.29 70	38.21 65	11.18 60	11.45 68	20.83 64

RMVSNet		15.61 69	22.67 43	10.90 71	13.04 72	14.34 72	12.27 58	33.07 37	5.30 74

Pnet-eth		14.99 70	18.95 52	12.35 70	19.54 67	4.23 77	7.59 67	30.32 41	13.27 66

PMVS	26.43 22	13.46 71	7.26 72	17.59 69	14.19 71	30.85 69	10.42 62	4.11 72	7.73 70
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
metmvs_fine		12.14 72	14.21 60	10.76 72	8.78 73	13.01 74	10.35 63	18.07 59	10.49 67

unMVSmet		7.50 73	3.60 73	10.10 73	6.98 74	18.02 71	3.83 72	3.37 73	5.28 75

Cas-MVS_preliminary		6.17 74	2.14 75	8.86 74	4.86 75	13.44 73	3.08 75	1.20 76	8.27 69

test_1120		5.30 75	3.59 74	6.45 77	4.63 76	4.96 76	3.98 71	3.19 74	9.76 68

CMPMVS	48.56 21	4.84 76	0.02 78	8.06 75	1.22 78	22.96 70	0.03 78	0.00 78	0.00 78
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
confMetMVS		4.71 77	2.07 76	6.46 76	4.51 77	11.02 75	1.90 76	2.24 75	3.85 76

FADENet		0.32 78	0.12 77	0.45 78	0.50 79	0.77 78	0.16 77	0.09 77	0.08 77

dnet		0.00 79	0.00 79	0.00 79	0.00 80	0.00 79	0.00 79	0.00 78	0.00 78

UnsupFinetunedMVSNet					16.34 68

Low-resolution many-view benchmark