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

unMVSv1		97.23 1	97.65 2	96.96 4	96.94 9	98.74 1	95.80 3	99.50 1	95.19 3

DeepPCF-MVS	98.03 4	96.80 4	95.79 8	97.47 2	98.75 3	98.48 2	93.36 7	98.23 5	95.19 3

PCF-MVS	98.23 3	97.14 2	96.43 4	97.61 1	99.41 1	98.23 3	95.62 4	97.24 14	95.20 2
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMP	97.00 8	96.57 7	96.40 5	96.68 6	97.45 5	98.22 4	94.57 5	98.23 5	94.36 7
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
DeepC-MVS_fast	98.92 1	96.93 3	96.34 6	97.33 3	98.83 2	98.21 5	93.89 6	98.79 3	94.96 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)		95.99 8	94.88 11	96.72 5	97.34 6	98.20 6	91.93 11	97.84 9	94.64 6

ACMH+	96.20 13	95.49 11	94.61 12	96.08 9	97.07 8	97.95 7	91.77 12	97.44 11	93.22 12

ACMM	97.17 6	96.69 5	97.35 3	96.24 8	97.15 7	97.93 8	96.54 2	98.16 7	93.65 10
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
IB-MVS	96.24 12	93.10 20	94.00 16	92.50 22	92.76 25	97.79 9	91.65 13	96.35 21	86.93 24
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	96.25 11	95.02 13	94.18 14	95.57 12	96.81 12	97.62 10	91.51 14	96.86 16	92.28 15
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PLC	98.56 2	95.81 9	95.80 7	95.82 11	95.90 16	97.46 11	93.21 8	98.39 4	94.11 9
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
COLMAP(base)		95.66 10	95.29 9	95.90 10	96.87 10	97.33 12	92.61 9	97.97 8	93.51 11

A1Net		94.08 16	92.17 23	95.34 13	96.68 13	97.20 13	87.75 24	96.60 20	92.15 17

A-TVSNet + Gipuma		94.87 14	95.11 10	94.71 15	95.40 18	97.17 14	92.41 10	97.81 10	91.56 19

DeepC-MVS	97.84 5	95.05 12	93.14 20	96.32 7	97.72 4	97.02 15	89.53 22	96.74 17	94.22 8
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
PVSNet_0	93.57 19	86.78 29	79.85 38	91.40 27	90.80 28	96.34 16	81.38 33	78.33 42	87.07 23

3Dnovator	95.63 14	87.41 26	80.54 34	92.00 23	96.84 11	96.13 17	78.30 39	82.78 37	83.03 29

3Dnovator+	95.58 15	87.31 27	80.52 35	91.84 25	96.45 14	96.11 18	77.89 40	83.15 36	82.96 31

TAPA-MVS(SR)		93.88 17	93.60 18	94.06 17	93.96 22	96.07 19	90.58 17	96.63 19	92.16 16

OpenMVS	95.20 17	86.65 30	79.07 39	91.71 26	96.24 15	95.92 20	77.14 42	81.00 41	82.97 30

COLMAP_ROB	97.10 7	94.23 15	94.16 15	94.27 16	95.49 17	95.89 21	91.39 15	96.93 15	91.43 20
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
test_mvsss		89.03 25	87.86 28	89.81 28	89.81 30	95.40 22	79.81 34	95.90 22	84.23 26

TAPA-MVS	96.40 10	93.41 18	93.58 19	93.30 18	92.79 24	95.34 23	90.46 18	96.70 18	91.77 18
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
PVSNet	94.91 18	84.37 35	76.88 44	89.36 30	89.15 31	95.01 24	81.51 32	72.25 51	83.92 27

CIDER		82.80 38	73.82 46	88.78 32	91.08 27	94.80 25	76.99 43	70.65 54	80.46 33
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
CPR_FA		90.55 23	88.48 27	91.92 24	90.55 29	94.73 26	87.56 25	89.40 30	90.48 22

tm-dncc		96.60 6	98.77 1	95.16 14	94.86 20	94.57 27	98.54 1	98.99 2	96.07 1

tmmvs		85.21 33	78.69 40	89.56 29	94.86 20	94.57 27	75.94 45	81.44 39	79.26 35

HY-MVS	96.53 9	89.40 24	89.89 25	89.07 31	87.26 34	94.23 29	86.76 26	93.03 27	85.74 25

GSE		93.24 19	94.32 13	92.51 21	92.17 26	94.21 30	91.39 15	97.25 13	91.15 21

Snet		71.04 54	50.26 75	84.90 37	81.53 46	94.02 31	46.31 74	54.21 77	79.14 36

test_robustmvs				77.12 49	88.60 32	93.21 32	82.52 31		49.55 72

test_1205		85.92 32	86.91 29	85.25 35	87.23 35	92.99 33	82.56 30	91.26 29	75.54 40

OpenMVS_ROB	88.34 20	79.83 40	71.69 49	85.26 34	87.13 36	92.62 34	72.32 50	71.05 53	76.04 39

LTVRE_ROB	95.29 16	92.81 21	92.60 22	92.96 19	93.47 23	92.40 35	89.87 20	95.32 24	93.01 13
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_fast		70.97 55	50.67 74	84.50 39	82.73 43	91.73 36	46.07 77	55.28 76	79.04 37

test_1126		82.97 36	81.27 32	84.10 40	87.91 33	91.40 37	77.35 41	85.20 32	72.99 45

R-MVSNet		86.00 31	85.66 30	86.23 33	84.90 37	90.99 38	85.10 28	86.21 31	82.78 32

BP-MVSNet		92.81 21	92.74 21	92.86 20	95.28 19	90.85 39	89.66 21	95.81 23	92.46 14
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
PVSNet_LR		82.84 37	80.11 36	84.66 38	83.98 39	90.58 40	76.03 44	84.20 33	79.43 34

MVS_test_1		75.95 46	72.76 48	78.07 46	78.02 49	90.51 41	61.59 59	83.93 34	65.68 57

MVE	68.59 21	71.88 51	69.57 51	73.41 59	66.25 66	88.47 42	67.60 55	71.54 52	65.52 59
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
LPCS		87.27 28	90.71 24	84.98 36	83.44 40	87.91 43	87.85 23	93.57 26	83.58 28

CasMVSNet(SR_A)		68.03 61	56.91 67	75.44 54	76.18 53	86.99 44	49.31 71	64.52 64	63.16 62

example		71.40 52	62.28 59	77.48 47	80.61 47	86.65 45	57.51 61	67.05 60	65.18 60

unsupervisedMVS_cas		79.34 41	77.46 43	80.59 41	82.05 44	86.11 46	78.48 37	76.43 46	73.62 44

hgnet		74.50 48	68.37 53	78.58 43	83.21 41	85.96 47	68.21 53	68.54 55	66.57 54

DPSNet		74.50 48	68.37 53	78.58 43	83.21 41	85.96 47	68.21 53	68.54 55	66.57 54

metmvs_fine		80.20 39	80.57 33	79.95 42	79.03 48	85.63 49	78.78 36	82.37 38	75.20 41

RMVSNet		77.00 44	82.85 31	73.11 60	77.13 50	85.46 50	74.25 47	91.44 28	56.73 70

CasMVSNet(base)		66.90 64	55.99 69	74.18 56	75.37 54	84.67 51	46.93 73	65.06 62	62.49 63

Pnet-new-		67.47 62	54.64 72	76.02 53	73.83 56	82.59 52	49.74 70	59.53 70	71.64 49

ANet-0.75		84.53 34	93.78 17	78.36 45	76.21 51	82.53 53	90.17 19	97.39 12	76.33 38

ANet		77.67 42	80.06 37	76.08 52	76.21 51	82.53 53	79.07 35	81.06 40	69.50 52

firsttry		70.92 56	69.37 52	71.96 61	70.42 62	80.33 55	64.15 58	74.59 48	65.15 61

mvs_zhu_1030		71.20 53	66.84 56	74.10 57	70.30 63	80.28 56	60.09 60	73.60 49	71.71 48

PSD-MVSNet		77.01 43	78.31 41	76.15 50	75.37 54	79.17 57	78.36 38	78.25 43	73.92 42

Pnet-blend++		66.07 65	59.38 61	70.53 62	73.22 58	79.12 58	50.55 68	68.21 57	59.25 65

Pnet-blend		66.07 65	59.38 61	70.53 62	73.22 58	79.12 58	50.55 68	68.21 57	59.25 65

vp_mvsnet		56.54 73	47.12 78	62.82 70	61.74 70	78.50 60	38.70 82	55.54 75	48.23 74

MVSNet_++		68.46 59	55.01 71	77.42 48	81.98 45	78.45 61	32.64 84	77.38 44	71.83 47

SGNet		75.22 47	75.91 45	74.77 55	73.83 56	77.61 62	74.98 46	76.83 45	72.87 46

test3		73.46 50	73.44 47	73.47 58	72.10 60	76.90 63	71.62 51	75.25 47	71.43 50

AttMVS		68.12 60	66.47 57	69.22 65	65.33 68	75.76 64	69.66 52	63.29 65	66.57 54

QQQNet		64.15 68	67.54 55	61.89 72	71.25 61	75.49 65	72.33 49	62.75 67	38.94 80

test_1124		56.89 72	47.96 77	62.84 69	53.21 81	74.83 66	39.90 80	56.03 74	60.47 64

MVSCRF		65.50 67	62.71 58	67.35 67	68.68 65	74.27 67	66.41 56	59.02 72	59.11 68

PMVS	60.66 23	47.91 80	37.76 82	54.68 77	58.99 78	73.94 68	46.15 76	29.37 83	31.10 82
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
TVSNet		69.80 57	69.63 50	69.92 64	69.08 64	72.29 69	65.78 57	73.47 50	68.39 53

P-MVSNet		76.80 45	89.15 26	68.57 66	65.42 67	70.78 70	84.62 29	93.68 25	69.51 51

MVSNet + Gipuma		60.80 69	57.55 66	62.97 68	61.50 71	70.61 71	52.16 66	62.94 66	56.80 69

F/T MVSNet+Gipuma		57.20 71	49.16 76	62.57 71	61.05 73	70.37 72	46.27 75	52.04 78	56.29 71

MVSNet_plusplus		69.70 58	60.08 60	76.11 51	84.56 38	70.03 73	53.06 65	67.09 59	73.74 43

CCVNet		56.43 74	56.46 68	56.42 76	60.81 75	69.50 74	53.44 64	59.47 71	38.94 80

unMVSmet		52.48 78	45.67 80	57.03 75	56.69 79	69.24 75	39.82 81	51.52 79	45.15 75

confMetMVS		48.96 79	45.93 79	50.97 81	44.11 82	67.03 76	41.02 79	50.84 80	41.78 77

MVSNet		59.21 70	55.77 70	61.50 73	61.91 69	63.44 77	46.98 72	64.56 63	59.16 67

CasMVSNet(SR_B)		55.71 75	58.78 63	53.66 78	53.67 80	62.71 78	51.51 67	66.05 61	44.59 76

FADENet		31.23 83	31.77 83	30.87 84	17.26 85	59.54 79	30.66 85	32.88 82	15.80 84

SVVNet		54.76 76	58.44 64	52.30 79	60.30 76	55.64 80	54.12 62	62.75 67	40.96 78

ternet		54.76 76	58.44 64	52.30 79	60.30 76	55.64 80	54.12 62	62.75 67	40.96 78

Pnet-eth		67.00 63	77.95 42	59.70 74	61.15 72	52.38 82	72.56 48	83.33 35	65.56 58

Cas-MVS_preliminary		43.58 81	50.71 73	38.83 83	34.27 83	51.77 83	43.48 78	57.95 73	30.45 83

test_1120		42.17 82	40.52 81	43.26 82	32.91 84	47.96 84	33.26 83	47.78 81	48.91 73

CMPMVS	66.12 22	7.25 84	0.67 84	11.64 85	5.59 86	29.32 85	1.34 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							85.16 27

UnsupFinetunedMVSNet					61.05 73

Low-resolution many-view benchmark