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

PCF-MVS	94.20 5	91.84 1	88.06 5	94.35 1	97.02 1	94.34 3	86.53 3	89.59 7	91.69 1
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
DeepPCF-MVS	95.94 2	91.68 2	88.50 3	93.80 2	95.92 2	94.67 1	86.12 4	90.88 4	90.82 3

DeepC-MVS_fast	96.59 1	91.26 3	88.40 4	93.17 3	95.61 3	93.73 4	85.46 5	91.35 3	90.16 4
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
unMVSv1		91.14 4	90.64 1	91.47 9	91.61 15	94.36 2	87.77 2	93.52 1	88.43 10

ACMP	92.05 9	90.39 5	87.34 7	92.42 4	94.01 4	93.60 5	84.39 8	90.29 6	89.66 6
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
PLC	95.54 3	89.80 6	87.72 6	91.19 10	92.06 11	92.55 10	84.60 7	90.83 5	88.96 8
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMM	91.95 10	89.71 7	86.80 8	91.65 8	93.63 6	92.83 9	84.20 9	89.40 8	88.49 9
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
tm-dncc		89.57 8	90.17 2	89.16 15	88.60 22	87.49 27	88.19 1	92.16 2	91.40 2

ACMH+	89.98 16	89.10 9	85.23 13	91.68 7	93.53 7	93.23 6	82.59 14	87.86 12	88.27 12

COLMAP(base)		89.06 10	86.32 10	90.89 11	92.63 9	92.02 13	83.78 11	88.86 10	88.00 13

DeepC-MVS	94.51 4	88.75 11	83.82 16	92.04 5	93.93 5	92.43 11	80.69 19	86.95 14	89.75 5
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
COLMAP(SR)		88.58 12	83.73 17	91.82 6	93.24 8	93.18 7	81.45 16	86.00 18	89.04 7

A-TVSNet + Gipuma		88.20 13	85.16 14	90.22 13	90.92 17	92.03 12	82.34 15	87.99 11	87.71 14

IB-MVS	92.85 6	87.70 14	85.59 11	89.11 16	90.51 19	93.12 8	84.64 6	86.53 16	83.70 23
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	89.72 17	87.53 15	83.42 18	90.28 12	92.12 10	91.92 14	81.05 17	85.79 20	86.79 18
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
A1Net		87.32 16	83.22 20	90.06 14	91.70 14	91.57 15	79.60 22	86.84 15	86.91 17

GSE		87.23 17	86.54 9	87.69 21	88.12 25	87.85 26	83.83 10	89.26 9	87.10 16

TAPA-MVS	92.12 8	86.58 18	84.40 15	88.04 20	88.32 24	89.30 21	82.72 13	86.09 17	86.49 19
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
COLMAP_ROB	90.47 14	86.35 19	83.04 21	88.57 18	90.38 20	89.41 20	80.60 20	85.48 22	85.91 20
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
BP-MVSNet		86.34 20	83.42 18	88.28 19	91.09 16	85.39 33	80.91 18	85.94 19	88.37 11
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
TAPA-MVS(SR)		86.31 21	82.84 22	88.63 17	88.90 21	89.73 18	80.05 21	85.62 21	87.26 15

LTVRE_ROB	88.28 18	83.47 22	81.17 23	85.00 27	86.77 27	82.40 38	79.16 23	83.18 23	85.82 21
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
CPR_FA		80.46 23	73.52 27	85.08 26	84.54 30	86.95 31	73.72 25	73.33 30	83.74 22

3Dnovator+	91.53 11	80.05 24	70.66 30	86.30 22	91.87 12	90.19 17	68.91 31	72.42 31	76.85 29

3Dnovator	91.47 12	79.93 25	70.75 28	86.05 24	91.87 12	89.67 19	69.16 28	72.34 32	76.61 30

HY-MVS	92.50 7	79.80 26	74.62 26	83.25 29	82.71 31	88.39 25	71.29 26	77.94 27	78.64 25

OpenMVS	90.15 15	79.06 27	69.23 32	85.61 25	90.88 18	89.10 22	67.39 33	71.07 33	76.86 28

PVSNet_0	88.03 19	78.98 28	68.05 34	86.27 23	87.63 26	91.03 16	69.06 29	67.04 37	80.16 24

LPCS		78.32 29	78.58 24	78.14 33	76.69 39	80.09 42	77.12 24	80.05 26	77.66 26

test_mvsss		76.77 30	70.70 29	80.81 32	81.46 34	87.41 29	59.27 38	82.14 24	73.58 32

PVSNet	91.05 13	76.45 31	65.65 35	83.65 28	84.78 29	88.84 24	67.99 32	63.31 44	77.35 27

ANet-0.75		76.04 32	85.27 12	69.88 45	68.75 54	74.61 51	82.75 12	87.80 13	66.29 40

tmmvs		75.74 33	65.05 36	82.86 31	88.60 22	87.49 27	64.31 35	65.80 40	72.49 34

test_1205		74.55 34	70.03 31	77.56 36	80.81 35	86.32 32	64.38 34	75.69 29	65.55 42

CIDER		73.88 35	60.41 41	82.87 30	86.29 28	89.03 23	62.67 37	58.14 50	73.28 33
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
R-MVSNet		73.82 36	69.12 33	76.96 37	75.29 41	81.93 40	69.06 29	69.18 34	73.66 31

PVSNet_LR		71.16 37	62.69 39	76.81 39	76.28 40	82.01 39	57.79 43	67.59 35	72.14 35

OpenMVS_ROB	79.82 20	70.36 38	59.33 43	77.72 34	79.39 36	84.88 34	59.25 39	59.41 48	68.88 38

test_1126		69.64 39	58.82 45	76.85 38	82.15 33	83.66 37	57.76 44	59.88 47	64.74 44

PSD-MVSNet		66.70 40	63.11 38	69.10 46	69.93 50	70.59 56	59.14 40	67.08 36	66.77 39

P-MVSNet		66.47 41	75.18 25	60.66 61	57.01 67	63.30 68	70.25 27	80.10 25	61.67 50

MVS_test_1		64.74 42	56.02 48	70.55 43	72.17 45	83.95 36	47.53 55	64.51 42	55.54 56

ANet		64.13 43	59.01 44	67.54 48	68.75 54	74.61 51	58.63 42	59.39 49	59.27 52

unsupervisedMVS_cas		64.00 44	53.69 50	70.87 42	73.43 43	75.95 48	49.45 51	57.93 51	63.23 47

SGNet		63.38 45	59.50 42	65.97 55	65.85 57	67.49 59	54.77 47	64.23 43	64.56 45

metmvs_fine		63.17 46	61.09 40	64.55 57	69.27 51	64.10 67	57.33 45	64.85 41	60.29 51

test3		61.43 47	57.40 46	64.12 58	63.85 59	66.58 60	52.46 49	62.34 45	61.91 48

RMVSNet		60.64 48	64.97 37	57.75 67	66.96 56	64.59 62	52.87 48	77.07 28	41.71 71

Snet		60.27 49	34.21 74	77.64 35	74.60 42	87.40 30	27.62 74	40.80 71	70.92 36

MVSNet_++		59.98 50	42.10 59	71.90 41	77.57 37	72.83 53	17.22 84	66.97 38	65.28 43

MVE	53.74 22	59.00 51	52.89 51	63.08 59	57.69 66	77.42 44	49.26 52	56.52 52	54.13 58
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
Pnet_fast		58.66 52	32.49 75	76.10 40	73.36 44	84.26 35	25.62 77	39.36 73	70.69 37

mvs_zhu_1030		58.65 53	47.62 55	66.00 54	64.63 58	71.59 54	41.49 59	53.76 54	61.78 49

hgnet		58.64 54	47.21 56	66.25 52	72.10 46	77.08 45	48.12 53	46.31 67	49.58 63

DPSNet		58.64 54	47.21 56	66.25 52	72.10 46	77.08 45	48.12 53	46.31 67	49.58 63

AttMVS		58.11 56	54.35 49	60.61 62	55.14 68	67.78 58	55.66 46	53.04 55	58.92 53

CasMVSNet(SR_A)		57.37 57	40.97 64	68.31 47	70.31 48	78.64 43	33.05 70	48.88 58	55.97 55

firsttry		57.36 58	51.28 53	61.41 60	61.20 64	71.35 55	46.45 57	56.11 53	51.67 60

example		57.19 59	43.25 58	66.49 51	70.25 49	80.14 41	40.18 60	46.33 66	49.07 65

Pnet-new-		57.13 60	37.61 69	70.14 44	68.85 53	75.84 49	35.69 63	39.52 72	65.74 41

MVSNet_plusplus		56.85 61	41.05 63	67.38 49	76.99 38	61.30 71	31.32 72	50.79 56	63.87 46

TVSNet		56.81 62	52.30 52	59.82 63	61.77 63	61.38 70	44.63 58	59.96 46	56.30 54

CasMVSNet(base)		55.90 63	39.77 66	66.66 50	69.07 52	75.83 50	31.70 71	47.84 62	55.08 57

QQQNet		51.78 64	49.94 54	53.00 69	63.63 60	64.21 66	51.24 50	48.64 59	31.18 79

MVSCRF		50.17 65	37.81 68	58.40 64	59.46 65	64.47 63	39.25 61	36.37 75	51.28 61

Pnet-blend++		49.57 66	36.71 71	58.15 65	61.80 61	64.39 64	26.12 75	47.30 64	48.26 67

Pnet-blend		49.57 66	36.71 71	58.15 65	61.80 61	64.39 64	26.12 75	47.30 64	48.26 67

Pnet-eth		49.22 68	56.60 47	44.30 78	48.09 75	34.94 83	46.73 56	66.48 39	49.88 62

MVSNet		47.72 69	40.64 65	52.43 70	49.39 71	55.25 75	33.58 69	47.70 63	52.66 59

MVSNet + Gipuma		46.36 70	39.39 67	51.01 72	48.29 74	59.78 72	36.67 62	42.10 70	44.96 69

F/T MVSNet+Gipuma		44.76 71	35.84 73	50.71 73	48.30 72	59.41 73	34.33 65	37.35 74	44.42 70

test_1124		43.61 72	28.84 76	53.46 68	45.65 79	66.26 61	23.79 80	33.88 76	48.47 66

SVVNet		43.25 73	41.22 61	44.61 76	53.30 69	47.23 79	33.79 67	48.64 59	33.31 74

ternet		43.25 73	41.22 61	44.61 76	53.30 69	47.23 79	33.79 67	48.64 59	33.31 74

CCVNet		41.80 75	36.92 70	45.05 75	45.82 78	58.15 74	30.97 73	42.86 69	31.18 79

vp_mvsnet		41.60 76	26.64 79	51.57 71	46.38 77	70.03 57	22.15 81	31.13 78	38.31 72

CasMVSNet(SR_B)		41.24 77	41.69 60	40.94 80	39.07 80	51.37 77	34.20 66	49.17 57	32.40 78

PMVS	49.05 23	37.77 78	25.95 80	45.66 74	47.10 76	63.26 69	35.19 64	16.71 82	26.61 82
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
unMVSmet		36.43 79	27.74 78	42.22 79	38.81 81	54.71 76	25.22 79	30.25 79	33.14 76

confMetMVS		34.16 80	28.37 77	38.02 81	33.00 82	50.95 78	25.52 78	31.21 77	30.12 81

Cas-MVS_preliminary		28.51 81	25.55 81	30.48 82	25.15 83	45.09 81	21.60 82	29.50 80	21.20 83

test_1120		25.78 82	22.14 82	28.20 83	17.89 84	34.03 84	17.74 83	26.54 81	32.68 77

FADENet		13.68 83	9.84 83	16.25 84	8.23 85	36.19 82	11.22 85	8.45 83	4.31 84

CMPMVS	61.59 21	6.00 84	0.25 84	9.83 85	3.28 86	26.21 85	0.51 86	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
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							63.85 36

test_robustmvs				65.49 56	82.67 32	76.18 47	58.77 41		37.61 73

UnsupFinetunedMVSNet					48.30 72

Low-resolution many-view benchmark