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	93.56 1	86.37 1	81.38 2	89.69 2	91.68 2	90.34 1	79.19 2	83.57 3	87.05 2

PCF-MVS	89.78 5	86.23 2	80.61 3	89.98 1	92.21 1	89.86 2	79.15 3	82.07 5	87.86 1
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
DeepC-MVS_fast	93.52 2	84.79 3	78.95 6	88.68 3	90.85 3	88.97 4	76.51 6	81.39 8	86.21 5
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMP	87.39 10	84.24 4	78.65 8	87.96 4	88.90 5	88.89 5	74.85 11	82.45 4	86.10 6
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
unMVSv1		84.05 5	83.50 1	84.41 16	84.75 18	87.18 12	81.46 1	85.54 1	81.30 21

PLC	91.07 3	83.49 6	79.02 5	86.47 9	87.55 11	87.15 13	76.24 7	81.81 6	84.70 9
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
IB-MVS	89.43 6	82.99 7	77.76 9	86.47 9	88.47 8	88.98 3	77.08 4	78.43 14	81.96 19
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+	83.78 15	82.73 8	75.83 15	87.33 6	88.67 6	88.50 6	73.89 12	77.77 18	84.84 8

ACMM	86.95 13	82.58 9	76.68 13	86.52 8	87.75 10	87.28 10	72.76 15	80.59 10	84.51 11
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
A-TVSNet + Gipuma		82.55 10	76.88 11	86.32 11	86.73 13	87.81 9	72.89 14	80.86 9	84.44 12

DeepC-MVS	91.02 4	82.53 11	74.98 18	87.57 5	88.50 7	87.92 8	72.07 19	77.88 17	86.28 4
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
tm-dncc		82.32 12	80.42 4	83.58 19	81.28 27	82.48 26	77.02 5	83.83 2	86.97 3

COLMAP(base)		82.31 13	76.85 12	85.95 12	87.50 12	86.55 15	75.41 10	78.29 15	83.82 13

COLMAP(SR)		81.85 14	73.79 19	87.23 7	89.02 4	87.99 7	72.66 17	74.92 19	84.68 10

A1Net		81.22 15	75.74 16	84.87 14	85.61 16	86.60 14	72.67 16	78.82 12	82.39 17

BP-MVSNet		81.15 16	75.41 17	84.97 13	87.81 9	81.23 31	72.11 18	78.70 13	85.88 7
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
GSE		80.93 17	78.72 7	82.41 22	82.36 24	81.72 30	75.91 8	81.53 7	83.15 15

TAPA-MVS	87.50 9	80.68 18	75.95 14	83.83 18	84.25 21	84.68 19	73.62 13	78.28 16	82.57 16
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
ACMH	83.09 17	79.47 19	71.54 20	84.76 15	86.56 14	85.52 17	70.21 21	72.87 22	82.19 18
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
TAPA-MVS(SR)		78.74 20	70.70 23	84.11 17	84.55 19	84.37 20	67.20 24	74.20 20	83.39 14

COLMAP_ROB	82.69 18	78.27 21	71.49 21	82.79 20	83.46 22	83.41 24	69.96 22	73.03 21	81.51 20
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
LTVRE_ROB	81.71 19	74.79 22	71.44 22	77.02 31	78.37 30	73.56 43	70.22 20	72.67 23	79.12 22
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
3Dnovator+	87.72 8	74.01 23	62.76 27	81.51 23	85.71 15	85.55 16	61.51 26	64.01 30	73.28 27

PVSNet_0	83.28 16	73.62 24	60.09 31	82.64 21	85.11 17	87.22 11	59.21 29	60.96 34	75.59 24

3Dnovator	87.35 11	73.38 25	62.96 26	80.33 24	84.34 20	84.16 22	61.54 25	64.39 29	72.49 30

HY-MVS	88.56 7	72.41 26	62.56 28	78.98 27	79.20 29	83.22 25	58.72 31	66.40 27	74.52 25

OpenMVS	85.28 14	72.17 27	61.03 30	79.59 26	82.41 23	83.50 23	58.91 30	63.15 32	72.86 29

PVSNet	87.13 12	71.39 28	58.61 32	79.90 25	81.65 25	84.27 21	59.32 28	57.90 37	73.79 26

CPR_FA		71.34 29	61.72 29	77.76 29	76.15 33	79.96 34	60.48 27	62.95 33	77.16 23

LPCS		70.58 30	67.95 24	72.33 34	70.49 39	73.62 42	67.73 23	68.16 25	72.89 28

ANet-0.75		69.45 31	77.48 10	64.10 46	61.68 51	69.60 49	75.76 9	79.19 11	61.02 41

tmmvs		67.95 32	53.69 37	77.45 30	81.28 27	82.48 26	53.71 34	53.66 43	68.60 32

CIDER		67.82 33	51.67 39	78.59 28	81.40 26	84.83 18	52.81 35	50.54 47	69.54 31
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
test_mvsss		65.81 34	55.91 34	72.41 33	74.70 36	80.45 32	45.55 42	66.28 28	62.09 40

test_1205		65.70 35	56.79 33	71.64 37	74.86 35	81.89 29	49.86 39	63.72 31	58.17 45

OpenMVS_ROB	73.86 20	63.94 36	50.71 41	72.76 32	73.57 37	79.97 33	50.36 36	51.06 46	64.76 36

R-MVSNet		63.29 37	55.31 35	68.61 40	66.45 44	74.27 40	55.32 33	55.31 41	65.10 35

PVSNet_LR		62.94 38	49.73 42	71.75 36	71.11 38	76.61 38	43.79 44	55.66 40	67.54 33

PSD-MVSNet		60.65 39	54.96 36	64.44 44	65.36 46	65.69 55	49.87 38	60.04 35	62.27 39

test_1126		59.75 40	43.11 49	70.84 38	76.19 32	77.23 37	41.68 47	44.53 54	59.11 43

P-MVSNet		58.87 41	63.50 25	55.79 56	51.35 65	59.52 62	58.47 32	68.53 24	56.50 48

SGNet		56.81 42	50.90 40	60.74 51	62.11 50	61.27 59	44.65 43	57.16 38	58.84 44

MVS_test_1		56.76 43	45.41 46	64.33 45	67.23 43	78.33 35	39.26 50	51.56 44	47.43 57

MVSNet_++		55.13 44	35.61 56	68.15 41	75.11 34	68.50 53	11.93 83	59.29 36	60.82 42

test3		54.74 45	47.91 43	59.30 53	60.21 53	61.13 60	41.75 46	54.07 42	56.54 47

unsupervisedMVS_cas		54.14 46	40.65 52	63.13 47	63.75 48	69.37 51	33.91 57	47.39 49	56.26 49

ANet		54.11 47	43.47 47	61.21 49	61.68 51	69.60 49	41.77 45	45.17 53	52.35 53

Snet		53.94 48	26.41 72	72.29 35	69.83 40	81.92 28	19.51 74	33.31 68	65.12 34

AttMVS		51.70 49	46.70 44	55.04 59	48.03 67	62.78 57	47.24 40	46.15 51	54.32 51

CasMVSNet(SR_A)		51.60 50	32.28 63	64.48 43	66.36 45	74.08 41	25.77 67	38.79 60	53.01 52

Pnet_fast		51.22 51	23.37 75	69.79 39	67.40 42	78.24 36	15.90 76	30.84 72	63.72 37

Pnet-new-		50.72 52	27.58 68	66.15 42	64.59 47	71.56 44	27.24 62	27.92 74	62.31 38

mvs_zhu_1030		50.59 53	37.13 55	59.57 52	57.86 59	66.44 54	30.92 59	43.34 55	54.39 50

TVSNet		49.80 54	42.25 50	54.83 60	57.91 58	56.82 65	34.18 55	50.32 48	49.77 55

CasMVSNet(base)		49.64 55	30.94 65	62.11 48	63.47 49	71.04 47	24.30 71	37.57 63	51.83 54

MVE	44.00 22	49.53 56	41.59 51	54.82 61	49.03 66	69.81 48	36.17 54	47.01 50	45.62 59
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
MVSNet_plusplus		48.34 57	29.36 66	60.99 50	69.62 41	56.31 66	20.92 72	37.80 61	57.05 46

firsttry		48.32 58	39.93 54	53.90 62	53.09 63	63.57 56	34.00 56	45.86 52	45.05 60

RMVSNet		47.98 59	52.85 38	44.74 70	55.96 60	47.68 74	39.09 51	66.61 26	30.56 73

example		47.27 60	33.53 59	56.43 55	59.94 55	75.51 39	31.42 58	35.63 64	33.83 70

hgnet		46.66 61	33.97 57	55.13 57	59.48 56	71.40 45	36.55 52	31.38 69	34.50 68

DPSNet		46.66 61	33.97 57	55.13 57	59.48 56	71.40 45	36.55 52	31.38 69	34.50 68

metmvs_fine		46.00 63	45.56 45	46.29 69	53.05 64	45.70 76	39.94 49	51.18 45	40.12 64

QQQNet		45.34 64	40.26 53	48.73 66	60.02 54	58.55 63	39.99 48	40.53 56	27.62 76

MVSNet		40.88 65	32.15 64	46.71 68	40.86 72	50.77 73	26.51 64	37.78 62	48.50 56

MVSCRF		40.84 66	26.49 71	50.40 63	47.42 70	58.10 64	27.02 63	25.96 75	45.69 58

Pnet-blend++		39.83 67	24.24 73	50.22 64	54.22 61	55.24 67	14.25 79	34.23 65	41.21 62

Pnet-blend		39.83 67	24.24 73	50.22 64	54.22 61	55.24 67	14.25 79	34.23 65	41.21 62

MVSNet + Gipuma		37.61 69	28.66 67	43.58 72	39.30 73	53.28 70	26.38 65	30.95 71	38.16 65

SVVNet		37.34 70	32.93 60	40.27 74	47.52 68	44.02 78	25.34 69	40.53 56	29.29 74

ternet		37.34 70	32.93 60	40.27 74	47.52 68	44.02 78	25.34 69	40.53 56	29.29 74

Pnet-eth		37.00 72	43.12 48	32.92 79	39.11 75	21.96 84	30.33 60	55.90 39	37.68 66

F/T MVSNet+Gipuma		36.65 73	26.87 69	43.17 73	38.81 76	53.12 71	25.52 68	28.22 73	37.56 67

test_1124		36.46 74	19.17 77	47.98 67	41.35 71	60.36 61	15.94 75	22.39 76	42.24 61

CCVNet		34.56 75	26.72 70	39.79 76	39.28 74	52.47 72	19.64 73	33.80 67	27.62 76

CasMVSNet(SR_B)		34.04 76	32.49 62	35.07 78	32.66 80	45.95 75	26.11 66	38.87 59	26.61 78

vp_mvsnet		33.19 77	16.83 78	44.10 71	37.77 79	62.70 58	14.98 78	18.68 77	31.84 71

PMVS	41.42 23	31.23 78	20.08 76	38.66 77	38.19 78	53.59 69	28.97 61	11.19 82	24.20 79
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
unMVSmet		26.14 79	16.71 79	32.43 80	29.12 81	44.88 77	15.84 77	17.58 78	23.29 80

confMetMVS		22.01 80	15.55 80	26.32 81	21.89 82	38.61 81	13.95 81	17.14 79	18.45 82

Cas-MVS_preliminary		19.83 81	13.19 81	24.25 82	17.25 83	39.23 80	12.65 82	13.73 81	16.28 83

test_1120		16.76 82	12.65 82	19.50 83	11.72 84	23.76 83	10.31 84	14.98 80	23.04 81

FADENet		5.54 83	3.06 83	7.20 85	3.62 85	16.50 85	3.91 85	2.21 83	1.49 84

CMPMVS	58.40 21	5.41 84	0.12 84	8.94 84	2.23 86	24.58 82	0.25 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							50.36 36

test_robustmvs				58.94 54	77.56 31	68.56 52	46.73 41		30.70 72

UnsupFinetunedMVSNet					38.81 76

Low-resolution many-view benchmark