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	80.84 1	63.41 1	53.95 1	69.71 2	69.72 4	71.26 1	48.73 1	59.18 1	68.16 2

TAPA-MVS	73.13 9	58.67 6	52.34 3	62.89 11	63.68 10	63.46 11	45.85 4	58.83 2	61.53 11
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
DeepC-MVS	79.81 2	62.37 2	52.77 2	68.77 3	68.65 5	70.30 3	46.95 2	58.59 3	67.37 3
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepC-MVS_fast	79.65 3	62.24 3	50.96 4	69.75 1	69.91 3	71.08 2	46.04 3	55.88 4	68.27 1
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
PCF-MVS	73.52 7	57.06 7	48.10 6	63.03 9	66.85 7	62.36 13	43.32 8	52.89 5	59.86 14
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMP	74.13 6	56.41 8	44.60 12	64.28 6	61.46 14	65.81 7	37.07 18	52.13 6	65.56 5
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
3Dnovator	76.31 5	59.36 5	48.19 5	66.81 5	70.04 2	67.59 5	44.27 5	52.10 7	62.80 10

OpenMVS	72.83 10	56.18 10	45.66 9	63.19 8	64.67 8	65.13 8	39.60 14	51.71 8	59.76 15

3Dnovator+	77.84 4	59.53 4	47.18 7	67.76 4	70.63 1	67.68 4	43.34 7	51.01 9	64.98 6

ACMM	73.20 8	55.01 13	43.19 16	62.89 11	59.60 17	66.07 6	35.89 20	50.48 10	63.01 9
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
PLC	70.83 11	53.50 15	45.79 8	58.64 17	56.31 25	60.76 16	41.70 11	49.88 11	58.86 17
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
IB-MVS	68.01 14	49.19 20	42.00 18	53.98 27	58.50 19	51.98 30	35.85 21	48.16 12	51.46 27

COLMAP(base)		55.51 11	44.97 11	62.54 13	60.85 15	62.46 12	42.24 9	47.71 13	64.29 7

ACMH+	68.96 13	52.02 18	43.83 14	57.49 20	59.85 16	51.27 33	40.04 12	47.62 14	61.35 12

LTVRE_ROB	69.57 12	53.52 14	45.46 10	58.89 15	58.76 18	60.60 18	43.91 6	47.01 15	57.32 20
Andreas Kuhn, Heiko Hirschmüller, Daniel Scharstein, Helmut Mayer: A TV Prior for High-Quality Scalable Multi-View Stereo Reconstruction. International Journal of Computer Vision 2016
TAPA-MVS(SR)		55.07 12	43.33 15	62.90 10	67.41 6	57.81 21	39.94 13	46.71 16	63.49 8

COLMAP(SR)		56.21 9	44.16 13	64.25 7	63.87 9	62.07 14	41.86 10	46.45 17	66.80 4

COLMAP_ROB	66.92 16	52.32 17	42.45 17	58.89 15	56.18 26	61.09 15	38.61 15	46.28 18	59.41 16
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
AttMVS		45.85 25	39.26 21	50.25 34	49.36 35	51.75 32	34.83 22	43.70 19	49.63 30

GSE		52.52 16	40.81 19	60.34 14	61.71 13	58.23 20	38.13 17	43.48 20	61.07 13

CIDER		47.21 23	38.36 23	53.11 28	55.37 28	55.10 28	33.25 26	43.46 21	48.87 32
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
LPCS		50.60 19	40.43 20	57.38 21	58.37 20	56.00 25	38.14 16	42.71 22	57.77 18

ACMH	67.68 15	47.97 22	38.24 24	54.45 26	56.88 22	50.08 34	34.49 23	41.99 23	56.40 21
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
OpenMVS_ROB	64.09 18	48.56 21	38.68 22	55.15 24	61.90 12	49.65 35	36.02 19	41.33 24	53.89 24

P-MVSNet		44.46 30	37.09 25	49.37 36	49.27 36	49.30 36	34.35 24	39.83 25	49.54 31

PVSNet_0	57.27 19	46.73 24	34.96 27	54.57 25	55.50 27	51.93 31	31.22 27	38.71 26	56.27 22

PVSNet	64.34 17	45.78 26	36.40 26	52.03 29	52.49 30	49.20 37	34.18 25	38.61 27	54.41 23

BP-MVSNet		43.22 31	32.08 28	50.65 33	52.86 29	46.16 43	27.25 29	36.92 28	52.94 25
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
R-MVSNet		36.87 39	29.78 29	41.61 42	42.00 42	46.99 39	24.73 30	34.83 29	35.83 42

test_1205		40.46 34	29.35 30	47.87 37	47.88 38	56.48 24	24.37 31	34.32 30	39.23 41

MVSNet		38.33 37	26.99 33	45.89 38	40.49 43	57.57 22	20.52 36	33.47 31	39.63 40

CasMVSNet(SR_B)		32.16 40	26.80 34	35.73 44	34.57 44	46.99 39	20.55 35	33.05 32	25.63 49

test_1126		45.72 27	28.57 32	57.15 22	62.33 11	59.06 19	24.22 32	32.91 33	50.08 29

CasMVSNet(SR_A)		45.13 28	25.99 35	57.89 18	56.32 24	64.81 9	19.76 38	32.23 34	52.54 26

MVSNet_++		25.72 46	17.38 46	31.27 46	33.47 45	18.28 57	2.87 65	31.90 35	42.06 38

CasMVSNet(base)		44.49 29	24.93 36	57.53 19	56.38 23	64.76 10	18.64 40	31.23 36	51.43 28

A-TVSNet + Gipuma		42.12 33	29.09 31	50.80 31	48.28 37	56.53 23	29.14 28	29.04 37	47.59 33

A1Net		23.74 48	23.95 38	23.61 49	21.43 53	19.62 54	19.93 37	27.96 38	29.78 45

CPR_FA		23.28 49	23.62 39	23.05 50	19.33 54	23.00 51	19.34 39	27.89 39	26.81 48

ANet		39.53 35	24.60 37	49.48 35	49.42 34	55.75 26	22.08 34	27.13 40	43.25 36

ANet-0.75		39.47 36	22.61 40	50.71 32	49.43 33	55.75 26	18.21 41	27.01 41	46.96 34

Pnet-eth		16.08 56	18.30 42	14.59 58	21.78 50	7.07 64	9.66 56	26.95 42	14.91 55

Pnet_fast		37.64 38	17.70 44	50.93 30	49.72 32	60.73 17	10.59 55	24.81 43	42.34 37

Pnet-blend		31.51 42	15.02 49	42.51 40	47.41 39	46.50 41	6.99 59	23.05 44	33.62 43

Pnet-blend++		31.51 42	15.02 49	42.51 40	47.41 39	46.50 41	6.99 59	23.05 44	33.62 43

Snet		23.85 47	17.41 45	28.15 47	32.72 46	24.90 50	12.22 51	22.60 46	26.83 47

MVSCRF		28.32 45	18.18 43	35.09 45	32.16 47	44.37 45	14.66 44	21.69 47	28.73 46

Pnet-new-		42.61 32	22.17 41	56.24 23	58.21 21	53.03 29	24.08 33	20.25 48	57.48 19

MVSNet_plusplus		28.64 44	12.93 55	39.11 43	49.80 31	21.55 52	6.17 61	19.69 49	45.98 35

MVE	26.22 22	16.26 55	16.97 47	15.79 57	11.75 60	19.40 55	14.45 45	19.48 50	16.21 53
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
metmvs_fine		10.03 61	14.31 52	7.18 65	6.75 65	6.98 65	9.52 57	19.10 51	7.81 59

RMVSNet		11.09 60	15.49 48	8.17 64	9.16 61	10.05 61	12.57 50	18.40 52	5.29 64

F/T MVSNet+Gipuma		17.31 51	14.39 51	19.26 52	16.83 56	19.67 53	14.03 46	14.75 53	21.27 50

MVSNet + Gipuma		16.91 52	14.12 53	18.77 55	16.39 58	19.13 56	13.98 47	14.27 54	20.78 51

firsttry		15.88 57	12.57 56	18.08 56	17.80 55	17.14 58	11.54 53	13.61 55	19.31 52

unMVSv1		13.05 59	12.03 59	13.73 59	14.18 59	16.18 59	12.13 52	11.93 56	10.84 58

hgnet		16.45 53	12.31 57	19.21 53	21.51 51	30.31 48	12.80 48	11.82 57	5.81 61

DPSNet		16.45 53	12.31 57	19.21 53	21.51 51	30.31 48	12.80 48	11.82 57	5.81 61

test_1124		31.65 41	13.28 54	43.90 39	44.21 41	47.60 38	14.80 43	11.75 59	39.88 39

example		15.69 58	10.14 61	19.39 51	23.73 49	30.32 47	11.10 54	9.17 60	4.13 65

PMVS	37.38 21	21.09 50	11.49 60	27.48 48	24.09 48	44.44 44	15.98 42	7.01 61	13.92 57
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
test_1120		9.53 62	6.82 62	11.33 61	8.74 62	9.06 62	7.57 58	6.06 62	16.20 54

unMVSmet		7.42 63	5.82 63	8.48 63	7.60 63	11.29 60	5.84 62	5.81 63	6.56 60

confMetMVS		4.86 66	3.39 65	5.85 66	5.59 66	6.61 66	2.94 64	3.83 64	5.35 63

Cas-MVS_preliminary		7.24 65	3.51 64	9.72 62	7.03 64	7.46 63	4.87 63	2.16 65	14.65 56

FADENet		0.17 67	0.15 66	0.18 67	0.31 68	0.16 67	0.19 66	0.10 66	0.08 66

dnet		0.00 68	0.00 68	0.00 68	0.00 69	0.00 68	0.00 68	0.00 67	0.00 67

CMPMVS	51.72 20	7.38 64	0.03 67	12.27 60	2.37 67	34.46 46	0.06 67	0.00 67	0.00 67
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
UnsupFinetunedMVSNet					16.83 56

Low-resolution many-view benchmark