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	89.82 1	78.32 1	71.76 2	82.70 2	83.10 3	84.81 2	69.51 3	74.00 1	80.18 4

PCF-MVS	84.09 5	77.67 2	71.71 3	81.63 5	82.20 6	82.75 7	70.09 2	73.34 4	79.95 6
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
IB-MVS	85.34 4	77.33 3	68.30 6	83.36 1	85.31 1	84.85 1	67.53 5	69.06 13	79.91 7
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
DeepC-MVS_fast	89.06 2	76.06 4	67.53 10	81.74 4	82.45 5	83.16 5	65.25 7	69.81 12	79.61 8
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMP	81.66 11	76.05 5	67.41 11	81.81 3	81.04 9	83.43 3	63.32 13	71.50 5	80.98 2
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
A-TVSNet + Gipuma		75.66 6	68.55 5	80.40 10	81.03 10	81.63 10	63.60 11	73.50 3	78.54 12

BP-MVSNet		74.96 7	65.37 14	81.36 8	84.33 2	77.41 23	60.79 18	69.96 10	82.34 1
Christian Sormann, Patrick KnÃ¶belreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
DeepC-MVS	86.58 3	74.68 8	65.20 16	80.99 9	79.97 11	82.67 8	62.18 17	68.23 15	80.34 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
ACMH+	76.62 16	74.57 9	64.12 18	81.55 6	81.57 8	83.05 6	63.23 14	65.00 18	80.01 5

PLC	83.97 7	74.32 10	67.97 7	78.55 13	78.24 16	79.96 14	65.84 6	70.10 9	77.45 14
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
A1Net		74.27 11	67.95 8	78.49 14	78.74 13	81.29 12	65.09 8	70.82 8	75.43 19

COLMAP(SR)		74.14 12	63.14 19	81.47 7	82.89 4	82.08 9	62.84 15	63.44 19	79.45 9

ACMM	80.70 13	73.45 13	64.34 17	79.53 11	77.78 18	81.48 11	59.83 19	68.84 14	79.34 10
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
TAPA-MVS	81.61 12	73.42 14	66.13 12	78.28 17	78.25 15	79.43 17	62.31 16	69.94 11	77.17 15
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
COLMAP(base)		73.14 15	65.21 15	78.42 16	78.47 14	79.75 16	64.64 9	65.79 17	77.05 16

unMVSv1		72.67 16	73.43 1	72.17 27	72.83 24	76.76 26	73.00 1	73.86 2	66.92 30

tm-dncc		72.53 17	67.61 9	75.81 20	71.97 26	76.76 26	63.80 10	71.42 6	78.69 11

GSE		71.29 18	65.66 13	75.04 23	72.79 25	75.92 30	63.54 12	67.78 16	76.41 17

TAPA-MVS(SR)		69.17 19	55.20 24	78.48 15	78.92 12	78.94 19	51.99 25	58.42 22	77.59 13

ACMH	75.40 17	69.16 20	56.85 22	77.36 18	77.71 19	78.92 20	56.51 22	57.18 23	75.47 18
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PVSNet_0	77.72 15	68.19 21	52.25 27	78.82 12	82.05 7	83.19 4	49.53 28	54.96 27	71.23 21

COLMAP_ROB	73.24 19	67.25 22	58.08 21	73.37 25	71.58 30	75.66 31	57.07 21	59.08 21	72.89 20
Johannes L. SchÃ¶nberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
3Dnovator+	82.88 8	66.67 23	53.47 25	75.47 21	77.63 20	80.17 13	52.43 24	54.50 30	68.61 26

PVSNet	82.34 9	66.30 24	51.74 30	76.01 19	77.97 17	79.90 15	50.71 27	52.76 34	70.17 23

HY-MVS	84.06 6	65.91 25	52.17 28	75.07 22	75.45 22	78.63 21	48.02 31	56.32 24	71.14 22

3Dnovator	82.32 10	65.50 26	53.26 26	73.66 24	75.04 23	78.44 22	51.58 26	54.95 28	67.50 28

OpenMVS	79.58 14	63.96 27	51.65 31	72.17 27	71.79 28	77.18 25	48.81 30	54.48 31	67.54 27

LTVRE_ROB	73.68 18	63.33 28	58.66 20	66.43 34	66.54 35	63.97 51	57.79 20	59.54 20	68.80 25
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
ANet-0.75		62.75 29	69.51 4	58.24 45	53.44 54	65.29 47	67.87 4	71.15 7	56.00 38

CPR_FA		61.93 30	52.13 29	68.46 30	65.68 38	70.87 36	49.19 29	55.06 26	68.83 24

LPCS		61.90 31	56.04 23	65.81 35	62.76 40	67.37 43	56.05 23	56.03 25	67.30 29

CIDER		61.36 32	43.94 34	72.97 26	75.61 21	79.14 18	43.66 33	44.21 42	64.16 31
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
tmmvs		59.72 33	43.26 36	70.69 29	71.97 26	76.76 26	43.36 34	43.16 43	63.34 32

OpenMVS_ROB	68.52 20	57.72 34	42.39 38	67.94 31	67.79 34	75.29 32	42.15 35	42.63 46	60.73 34

test_1205		56.70 35	43.64 35	65.41 36	67.82 33	76.71 29	35.46 42	51.82 35	51.71 45

PVSNet_LR		55.55 36	38.58 44	66.86 32	65.81 37	72.43 34	32.04 45	45.12 41	62.35 33

PSD-MVSNet		55.32 37	48.24 33	60.05 43	61.77 42	61.23 53	42.13 36	54.34 32	57.14 37

test_mvsss		53.10 38	39.96 41	61.86 40	65.92 36	71.33 35	32.70 43	47.23 38	48.34 52

R-MVSNet		52.36 39	41.69 39	59.48 44	56.87 51	66.76 44	40.62 37	42.75 44	54.80 40

SGNet		51.07 40	43.18 37	56.33 48	58.93 47	56.69 59	36.19 41	50.16 36	53.35 42

P-MVSNet		50.96 41	50.55 32	51.23 55	46.21 60	56.33 60	46.98 32	54.13 33	51.14 46

test_1126		50.94 42	30.98 52	64.25 37	68.59 32	70.82 37	27.99 49	33.96 55	53.35 42

MVSNet_++		50.11 43	29.32 55	63.96 38	71.79 28	64.50 49	9.07 79	49.58 37	55.59 39

test3		49.10 44	39.66 42	55.39 50	57.08 49	57.23 57	32.44 44	46.89 39	51.85 44

Snet		48.52 45	21.09 66	66.81 33	65.31 39	77.32 24	14.23 72	27.95 63	57.79 36

MVS_test_1		48.01 46	35.49 45	56.36 47	60.38 44	70.57 39	30.45 47	40.53 47	38.12 59

CasMVSNet(SR_A)		46.94 47	25.61 60	61.16 42	61.92 41	70.76 38	20.15 64	31.08 60	50.81 47

unsupervisedMVS_cas		45.88 48	30.86 53	55.89 49	53.72 53	64.16 50	23.38 58	38.35 49	49.80 48

AttMVS		45.75 49	39.65 43	49.81 57	41.73 66	58.57 56	39.10 38	40.19 48	49.14 51

ANet		45.65 50	31.83 49	54.86 51	53.44 54	65.29 47	28.83 48	34.84 53	45.84 54

Pnet-new-		45.10 51	20.14 68	61.74 41	59.27 46	67.89 41	20.77 61	19.51 75	58.07 35

Pnet_fast		44.79 52	17.59 73	62.93 39	61.48 43	72.87 33	10.49 76	24.68 67	54.43 41

CasMVSNet(base)		44.67 53	24.42 63	58.16 46	57.66 48	67.67 42	18.86 67	29.99 62	49.16 50

TVSNet		44.57 54	34.62 46	51.20 56	54.62 52	53.74 64	26.57 53	42.67 45	45.25 55

mvs_zhu_1030		42.95 55	28.13 56	52.84 53	49.44 56	61.49 52	22.04 59	34.22 54	47.58 53

firsttry		41.41 56	32.59 47	47.30 58	44.84 63	57.23 57	27.29 50	37.89 50	39.82 57

QQQNet		40.37 57	32.52 48	45.60 59	56.88 50	54.85 63	31.38 46	33.66 56	25.07 70

MVSNet_plusplus		40.24 58	19.67 70	53.95 52	59.78 45	52.38 66	13.77 73	25.56 66	49.68 49

MVE	35.65 22	39.97 59	31.66 50	45.51 60	39.04 68	60.77 54	25.59 55	37.73 51	36.74 61
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
example		36.83 60	25.33 62	44.50 61	46.40 59	68.80 40	24.57 56	26.08 65	18.29 78

hgnet		35.63 61	23.84 64	43.49 63	44.86 61	65.90 45	27.26 51	20.42 72	19.70 76

DPSNet		35.63 61	23.84 64	43.49 63	44.86 61	65.90 45	27.26 51	20.42 72	19.70 76

MVSNet		35.27 63	25.46 61	41.81 68	34.15 71	47.84 68	20.52 62	30.39 61	43.43 56

RMVSNet		33.89 64	40.18 40	29.69 77	40.15 67	31.89 78	25.80 54	54.56 29	17.03 79

MVSCRF		33.82 65	19.17 72	43.58 62	38.65 69	52.40 65	18.43 68	19.92 74	39.67 58

SVVNet		32.72 66	26.54 57	36.84 70	42.17 64	41.73 74	19.41 65	33.66 56	26.61 66

ternet		32.72 66	26.54 57	36.84 70	42.17 64	41.73 74	19.41 65	33.66 56	26.61 66

Pnet-blend++		32.27 68	16.31 74	42.91 66	47.22 57	47.07 71	8.20 80	24.42 68	34.45 62

Pnet-blend		32.27 68	16.31 74	42.91 66	47.22 57	47.07 71	8.20 80	24.42 68	34.45 62

test_1124		30.98 70	12.45 77	43.34 65	37.24 70	55.54 61	11.55 75	13.35 76	37.24 60

MVSNet + Gipuma		29.83 71	19.95 69	36.42 72	29.64 76	47.72 69	18.31 69	21.58 70	31.90 64

F/T MVSNet+Gipuma		29.40 72	19.29 71	36.14 73	29.17 77	47.69 70	18.06 70	20.51 71	31.56 65

CCVNet		29.33 73	20.37 67	35.30 74	32.96 72	47.86 67	13.64 74	27.11 64	25.07 70

CasMVSNet(SR_B)		28.66 74	26.02 59	30.42 76	27.87 80	41.07 76	20.42 63	31.61 59	22.30 74

metmvs_fine		28.59 75	29.67 54	27.87 78	30.47 75	29.64 79	24.16 57	35.17 52	23.51 73

vp_mvsnet		26.59 76	9.92 78	37.70 69	31.17 74	55.39 62	9.68 77	10.15 77	26.55 68

Pnet-eth		26.42 77	31.46 51	23.06 79	31.47 73	11.87 84	17.32 71	45.61 40	25.85 69

PMVS	34.80 23	24.26 78	14.57 76	30.71 75	28.60 79	43.77 73	21.54 60	7.60 79	19.76 75
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
unMVSmet		16.34 79	9.37 79	20.99 80	17.71 81	32.57 77	9.11 78	9.64 78	12.68 81

Cas-MVS_preliminary		12.65 80	6.00 82	17.09 81	9.38 83	29.38 80	7.06 82	4.93 82	12.51 82

confMetMVS		12.44 81	7.02 80	16.05 82	12.24 82	27.72 81	6.80 83	7.25 81	8.20 83

test_1120		10.10 82	6.98 81	12.17 83	7.50 84	13.03 83	6.62 84	7.35 80	15.99 80

CMPMVS	54.94 21	5.08 83	0.05 84	8.44 84	1.63 86	23.68 82	0.10 86	0.00 84	0.00 85
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
FADENet		1.77 84	0.82 83	2.40 85	1.65 85	5.07 85	1.06 85	0.57 83	0.47 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							38.64 39

test_robustmvs				51.41 54	71.54 31	58.89 55	37.31 40		23.80 72

UnsupFinetunedMVSNet					29.17 77

Low-resolution many-view benchmark