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

3Dnovator	92.54 3	87.14 1	82.53 4	90.22 2	92.38 6	87.90 7	76.07 3	88.99 3	90.37 3

test_1120		85.87 2	90.81 1	82.58 24	92.77 5	71.70 39	90.04 2	91.57 1	83.26 25

test_1124		85.71 3	89.49 2	83.20 19	93.00 4	69.67 43	90.66 1	88.31 4	86.91 14

3Dnovator+	92.74 2	85.26 4	79.20 5	89.30 6	91.06 11	85.88 12	71.70 7	86.71 5	90.95 1

LTVRE_ROB	93.87 1	84.10 5	76.06 10	89.46 5	92.03 7	89.46 3	74.68 5	77.44 25	86.88 16
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
Pnet-new-		83.71 6	82.83 3	84.29 17	91.11 10	71.44 41	75.11 4	90.55 2	90.33 4

MVSNet		83.40 7	73.23 16	90.18 3	96.40 3	96.26 1	60.66 19	85.80 7	77.89 38

tmmvs		83.24 8	76.41 9	87.80 9	91.55 8	83.34 17	70.47 8	82.35 14	88.50 11

OpenMVS	89.45 8	82.68 9	76.80 8	86.60 13	87.53 16	85.75 13	67.54 10	86.07 6	86.53 18

AttMVS		82.46 10	75.33 12	87.21 11	96.89 2	75.50 32	65.76 14	84.90 8	89.24 7

CasMVSNet(SR_B)		82.29 11	64.88 29	93.90 1	97.33 1	93.46 2	52.57 33	77.19 26	90.91 2

DeepC-MVS	91.39 4	81.68 12	75.54 11	85.78 14	85.54 23	84.87 15	66.52 11	84.56 10	86.91 14
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
test_1126		81.50 13	76.94 7	84.53 16	90.07 12	80.09 22	72.16 6	81.72 15	83.44 24

DeepPCF-MVS	90.46 6	81.47 14	72.84 17	87.22 10	86.87 18	85.94 11	64.61 17	81.08 18	88.86 9

mvs_zhu_1030		81.21 15	70.41 20	88.40 8	89.88 13	88.12 5	56.56 28	84.27 11	87.20 13

DeepC-MVS_fast	89.96 7	80.90 16	71.96 18	86.87 12	86.16 19	85.65 14	63.68 18	80.23 19	88.79 10
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
tm-dncc		80.42 17	68.45 23	88.41 7	91.55 8	83.34 17	58.49 24	78.40 24	90.33 4

TAPA-MVS	88.58 10	79.17 18	74.71 13	82.14 26	83.86 26	80.69 21	65.33 15	84.10 12	81.86 27
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
Pnet-blend++		79.10 19	73.55 14	82.79 22	85.61 21	87.39 8	65.83 12	81.27 16	75.38 40

Pnet-blend		79.10 19	73.55 14	82.79 22	85.61 21	87.39 8	65.83 12	81.27 16	75.38 40

TAPA-MVS(SR)		79.05 21	76.98 6	80.43 29	86.95 17	72.03 38	69.38 9	84.57 9	82.32 26

LPCS		77.52 22	66.30 27	85.00 15	84.90 24	83.49 16	59.63 22	72.97 34	86.62 17

COLMAP_ROB	91.06 5	77.39 23	68.75 22	83.16 20	77.98 32	82.59 19	57.72 26	79.78 20	88.89 8
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
P-MVSNet		77.22 24	69.44 21	82.40 25	88.76 15	72.48 37	59.92 21	78.95 22	85.96 19

COLMAP(base)		75.61 25	66.44 26	81.73 27	78.81 30	78.86 23	58.69 23	74.19 30	87.52 12

GSE		75.15 26	62.89 35	83.32 18	83.05 27	81.11 20	55.75 31	70.03 42	85.80 20

CasMVSNet(base)		74.99 27	62.90 34	83.05 21	79.99 29	88.10 6	49.26 38	76.54 27	81.07 30

OpenMVS_ROB	85.12 16	74.56 28	71.88 19	76.34 36	86.01 20	62.96 47	65.18 16	78.58 23	80.05 32

CasMVSNet(SR_A)		73.59 29	63.86 32	80.08 30	73.85 37	86.72 10	51.68 36	76.04 28	79.66 34

PCF-MVS	84.52 17	73.49 30	62.30 36	80.95 28	84.42 25	78.59 24	55.89 30	68.72 45	79.83 33
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
PLC	85.34 15	73.20 31	65.63 28	78.25 32	73.53 39	77.42 27	57.50 27	73.77 32	83.78 23
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
COLMAP(SR)		73.18 32	64.70 30	78.83 31	76.92 34	75.11 33	58.32 25	71.09 39	84.45 22

HY-MVS	82.50 18	72.16 33	66.92 25	75.64 37	78.74 31	69.74 42	60.03 20	73.81 31	78.45 37

ACMM	88.83 9	70.91 34	60.87 37	77.60 33	73.74 38	78.44 25	50.09 37	71.64 36	80.61 31
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
ACMP	88.15 13	70.01 35	59.11 40	77.27 34	72.09 42	77.94 26	48.49 39	69.73 43	81.78 28
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
ACMH+	88.43 11	69.12 36	64.25 31	72.37 40	75.42 36	62.09 48	55.28 32	73.23 33	79.59 35

ACMH	88.36 12	68.62 37	63.35 33	72.13 42	76.30 35	61.45 50	52.32 34	74.39 29	78.63 36
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PMVS	87.21 14	68.59 38	37.06 57	89.61 4	89.58 14	88.93 4	43.78 45	30.34 68	90.31 6
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
CIDER		68.12 39	67.69 24	68.40 44	67.10 46	67.39 44	56.32 29	79.06 21	70.71 46
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
vp_mvsnet		66.63 40	50.71 47	77.24 35	80.40 28	74.20 34	17.94 74	83.49 13	77.11 39

Pnet_fast		66.28 41	58.61 41	71.39 43	67.96 45	76.81 31	46.02 43	71.20 38	69.40 48

PVSNet_LR		65.63 42	55.66 42	72.28 41	69.26 44	73.17 35	40.76 47	70.56 41	74.42 42

ANet		64.16 43	49.90 48	73.66 38	72.65 40	77.24 28	43.69 46	56.11 49	71.09 44

IB-MVS	77.21 19	63.04 44	59.74 39	65.23 48	69.93 43	62.02 49	47.98 40	71.51 37	63.73 51
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
R-MVSNet		62.92 45	59.84 38	64.98 49	66.63 48	67.11 46	47.38 41	72.30 35	61.19 52

test_1205		60.45 46	51.48 45	66.42 45	66.80 47	73.02 36	46.79 42	56.17 48	59.45 54

PVSNet	76.22 20	60.21 47	55.24 43	63.53 50	62.99 52	56.76 52	51.86 35	58.62 46	70.83 45

PVSNet_0	70.34 21	60.14 48	51.55 44	65.87 47	64.30 51	59.84 51	45.76 44	57.35 47	73.47 43

MVSCRF		58.45 49	50.89 46	63.49 51	65.53 49	67.21 45	32.48 52	69.31 44	57.74 55

ANet-0.75		55.72 50	29.52 65	73.18 39	72.65 40	77.24 28	22.81 62	36.23 63	69.65 47

A-TVSNet + Gipuma		55.66 51	39.87 51	66.19 46	61.25 53	71.47 40	39.48 48	40.27 61	65.86 50

BP-MVSNet		54.95 52	44.08 49	62.20 52	64.58 50	55.08 54	35.81 49	52.35 52	66.94 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
MVSNet_plusplus		53.83 53	41.31 50	62.18 53	77.12 33	27.65 63	11.93 80	70.69 40	81.76 29

unsupervisedMVS_cas		45.33 54	35.07 58	52.17 54	49.98 56	55.86 53	26.75 58	43.39 56	50.68 56

Pnet-eth		41.50 55	37.56 54	44.12 56	59.98 54	35.12 60	25.76 59	49.36 53	37.26 64

MVS_test_1		41.26 56	33.59 60	46.37 55	57.31 55	36.32 58	14.34 79	52.84 51	45.48 57

F/T MVSNet+Gipuma		37.77 57	38.70 52	37.15 61	41.21 59	26.31 64	34.11 50	43.29 57	43.92 58

MVSNet_++		36.74 58	29.99 64	41.25 57	41.08 61	21.58 76	4.20 84	55.79 50	61.07 53

MVSNet + Gipuma		36.49 59	37.36 55	35.92 63	39.51 63	25.68 70	33.78 51	40.94 59	42.57 61

Snet		35.89 60	38.02 53	34.46 64	42.09 58	25.34 71	31.63 53	44.41 55	35.96 67

Cas-MVS_preliminary		34.22 61	23.72 72	41.22 58	24.86 69	13.53 81	21.84 66	25.60 77	85.27 21

CPR_FA		31.79 62	33.68 59	30.53 65	26.52 68	27.95 62	26.79 57	40.57 60	37.13 65

CCVNet		31.72 63	23.36 76	37.29 60	32.94 67	35.83 59	23.09 61	23.62 79	43.10 59

test_mvsss		29.61 64	19.32 79	36.47 62	33.95 66	44.00 55	10.69 81	27.94 72	31.45 70

hgnet		28.77 65	28.46 67	28.98 67	37.72 64	36.88 56	22.34 63	34.57 64	12.35 80

DPSNet		28.77 65	28.46 67	28.98 67	37.72 64	36.88 56	22.34 63	34.57 64	12.35 80

A1Net		28.59 67	31.33 62	26.77 73	23.32 71	20.44 79	25.11 60	37.56 62	36.53 66

unMVSmet		27.29 68	37.09 56	20.76 79	21.57 73	19.92 80	29.37 54	44.81 54	20.79 75

QQQNet		27.27 69	23.68 73	29.66 66	20.58 75	25.28 72	20.44 70	26.93 73	43.10 59

TVSNet		26.87 70	26.31 69	27.24 72	22.04 72	25.97 66	21.82 67	30.81 67	33.71 68

SVVNet		26.60 71	23.52 74	28.65 69	19.11 79	25.79 68	20.11 71	26.93 73	41.05 62

ternet		26.60 71	23.52 74	28.65 69	19.11 79	25.79 68	20.11 71	26.93 73	41.05 62

MVE	59.87 23	26.31 73	31.05 63	23.15 77	16.99 83	26.25 65	28.39 55	33.71 66	26.21 73
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
test3		25.90 74	25.56 70	26.12 74	20.86 74	25.97 66	21.87 65	29.25 71	31.54 69

CMPMVS	68.83 22	24.56 75	0.66 83	40.49 59	44.64 57	76.82 30	1.32 85	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
SGNet		23.30 76	22.41 77	23.89 76	19.97 77	24.65 73	18.80 73	26.01 76	27.05 72

example		22.96 77	15.83 82	27.71 71	39.92 62	34.87 61	16.64 77	15.02 82	8.35 82

RMVSNet		22.89 78	28.83 66	18.93 80	20.40 76	21.96 75	28.10 56	29.56 69	14.42 79

firsttry		22.43 79	18.70 80	24.92 75	24.56 70	21.07 77	17.26 75	20.14 80	29.14 71

PSD-MVSNet		22.04 80	20.75 78	22.90 78	19.67 78	23.59 74	17.26 75	24.24 78	25.46 74

metmvs_fine		20.53 81	32.00 61	12.88 83	12.30 84	11.56 83	21.14 68	42.86 58	14.77 78

confMetMVS		20.09 82	25.00 71	16.82 82	18.50 82	12.38 82	20.47 69	29.53 70	19.57 76

unMVSv1		17.38 83	16.06 81	18.27 81	19.03 81	20.72 78	15.92 78	16.19 81	15.05 77

FADENet		0.40 84	0.50 84	0.33 85	0.58 86	0.22 85	0.62 86	0.38 83	0.18 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							10.18 82

test_robustmvs				4.01 84	6.76 85	3.49 84	6.41 83		1.77 83

UnsupFinetunedMVSNet					41.21 59

Low-resolution many-view benchmark