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

DeepC-MVS_fast	98.69 1	96.77 1	95.30 1	97.74 1	98.24 2	97.85 3	92.74 2	97.86 1	97.13 1
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
DeepPCF-MVS	98.18 3	96.68 2	95.10 2	97.74 1	98.30 1	97.78 5	92.41 3	97.79 2	97.12 2

DeepC-MVS	98.35 2	96.09 3	94.20 11	97.34 3	97.94 4	97.39 11	90.85 14	97.56 3	96.71 4
Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer: DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction. 3DV 2020
COLMAP(SR)		96.07 4	94.39 8	97.19 5	96.86 8	97.92 2	91.54 10	97.24 8	96.79 3

TAPA-MVS(SR)		95.82 5	94.80 3	96.49 9	96.19 12	97.58 10	92.04 4	97.56 3	95.71 12

PCF-MVS	97.08 13	95.68 6	93.36 15	97.23 4	98.16 3	96.93 15	91.43 11	95.29 20	96.61 5
Andreas Kuhn, Shan Lin, Oliver Erdler: Plane Completion and Filtering for Multi-View Stereo Reconstruction. GCPR 2019
ACMM	97.58 5	95.68 6	94.49 6	96.48 10	96.12 13	97.20 14	92.04 4	96.94 11	96.10 9
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
ACMH+	97.24 9	95.61 8	93.77 14	96.83 6	96.69 10	97.71 7	90.70 15	96.85 12	96.09 10

TAPA-MVS	97.07 14	95.48 9	94.55 4	96.10 12	95.68 16	97.30 12	91.73 8	97.37 6	95.32 15
Andrea Romanoni, Matteo Matteucci: TAPA-MVS: Textureless-Aware PAtchMatch Multi-View Stereo. ICCV 2019
COLMAP(base)		95.42 10	94.03 13	96.35 11	95.96 14	96.92 16	91.10 13	96.97 9	96.16 8

PLC	97.94 4	95.35 11	94.28 10	96.06 13	95.26 20	96.59 17	91.61 9	96.96 10	96.34 7
Jie Liao, Yanping Fu, Qingan Yan, Chunxia xiao: Pyramid Multi-View Stereo with Local Consistency. Pacific Graphics 2019
ACMH	97.28 7	95.29 12	93.25 16	96.65 7	96.32 11	97.93 1	89.82 17	96.68 15	95.69 13
Qingshan Xu and Wenbing Tao: Multi-Scale Geometric Consistency Guided Multi-View Stereo. CVPR 2019
ACMP	97.20 10	95.11 13	93.01 19	96.52 8	95.68 16	97.26 13	89.74 18	96.27 17	96.61 5
Qingshan Xu and Wenbing Tao: Planar Prior Assisted PatchMatch Multi-View Stereo. AAAI 2020
LTVRE_ROB	97.16 11	95.10 14	94.13 12	95.75 14	95.69 15	95.73 21	92.87 1	95.39 19	95.83 11
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
GSE		95.00 15	94.33 9	95.45 17	94.81 21	96.53 19	91.98 6	96.68 15	95.02 17

A-TVSNet + Gipuma		94.96 16	94.54 5	95.24 19	95.31 19	96.58 18	91.82 7	97.27 7	93.83 18

BP-MVSNet		94.57 17	93.20 17	95.49 15	96.77 9	94.34 29	89.67 19	96.74 13	95.35 14
Christian Sormann, Patrick Knöbelreiter, Andreas Kuhn, Mattia Rossi, Thomas Pock, Friedrich Fraundorfer: BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo. 3DV 2020
COLMAP_ROB	97.56 6	94.43 18	93.09 18	95.33 18	94.55 22	96.37 20	89.45 20	96.72 14	95.07 16
Johannes L. Schönberger, Enliang Zheng, Marc Pollefeys, Jan-Michael Frahm: Pixelwise View Selection for Unstructured Multi-View Stereo. ECCV 2016
IB-MVS	95.67 17	94.34 19	94.41 7	94.29 22	95.45 18	97.77 6	91.37 12	97.44 5	89.67 26

3Dnovator	97.25 8	92.47 20	87.97 28	95.47 16	97.93 5	97.85 3	86.02 24	89.92 28	90.64 22

3Dnovator+	97.12 12	92.17 21	87.57 29	95.24 19	97.49 6	97.68 8	85.24 27	89.91 29	90.54 23

LPCS		92.07 22	92.93 20	91.49 27	90.32 31	93.29 31	90.66 16	95.20 21	90.87 21

OpenMVS	96.50 15	91.77 23	86.65 31	95.19 21	97.38 7	97.66 9	84.60 30	88.69 33	90.53 24

R-MVSNet		90.93 24	89.80 22	91.68 26	90.81 30	94.15 30	87.67 21	91.93 26	90.07 25

PVSNet_0	94.43 18	90.46 25	85.77 32	93.59 23	93.82 24	95.33 24	84.62 29	86.92 35	91.61 20

test_1126		89.66 26	88.06 27	90.73 29	93.40 25	95.29 25	84.99 28	91.13 27	83.50 34

test_1205		89.53 27	89.08 25	89.83 32	92.63 27	95.71 22	85.34 26	92.82 25	81.15 40

CPR_FA		89.35 28	88.37 26	90.00 31	89.38 33	87.69 39	87.29 23	89.46 31	92.93 19

PVSNet	96.02 16	89.22 29	84.53 36	92.35 24	93.29 26	94.60 28	85.85 25	83.20 38	89.17 27

CIDER		88.55 30	83.13 38	92.16 25	93.95 23	95.61 23	83.79 33	82.47 39	86.92 29
Qingshan Xu and Wenbing Tao: Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume. AAAI 2020
ANet-0.75		87.89 31	89.58 23	86.76 33	84.37 45	90.29 35	84.51 31	94.65 22	85.61 32

OpenMVS_ROB	92.34 19	87.46 32	81.75 39	91.27 28	92.55 28	95.21 26	81.21 37	82.29 40	86.05 30

P-MVSNet		85.35 33	91.57 21	81.21 42	78.76 52	82.87 47	87.56 22	95.58 18	82.00 37

ANet		85.16 34	85.20 34	85.14 36	84.37 45	90.29 35	82.00 35	88.39 34	80.78 42

unMVSv1		83.89 35	84.37 37	83.58 38	84.00 49	85.63 41	83.17 34	85.56 36	81.11 41

metmvs_fine		83.50 36	84.90 35	82.56 40	84.72 40	81.80 50	81.09 38	88.72 32	81.17 39

A1Net		82.90 37	89.39 24	78.57 51	85.19 37	63.18 64	84.36 32	94.42 23	87.34 28

RMVSNet		80.62 38	86.71 30	76.57 53	84.14 47	78.09 52	80.22 39	93.20 24	67.48 59

Pnet_fast		80.38 39	65.52 56	90.29 30	90.03 32	94.91 27	60.54 55	70.51 58	85.93 31

AttMVS		80.32 40	78.27 40	81.68 41	78.98 51	86.16 40	79.76 40	76.79 53	79.89 43

Pnet-new-		79.46 41	69.54 52	86.08 34	84.76 39	90.31 34	65.13 49	73.95 55	83.16 35

CasMVSNet(SR_A)		79.00 42	69.56 51	85.29 35	85.95 35	92.82 32	62.08 52	77.05 49	77.11 44

hgnet		78.82 43	76.20 42	80.57 46	84.50 41	82.32 48	75.56 42	76.84 50	74.88 48

DPSNet		78.82 43	76.20 42	80.57 46	84.50 41	82.32 48	75.56 42	76.84 50	74.88 48

Pnet-blend++		78.13 45	73.62 46	81.13 43	84.46 43	88.20 37	66.50 46	80.74 42	70.74 57

Pnet-blend		78.13 45	73.62 46	81.13 43	84.46 43	88.20 37	66.50 46	80.74 42	70.74 57

CasMVSNet(base)		78.10 47	68.51 53	84.50 37	85.47 36	91.56 33	59.79 57	77.22 48	76.47 45

Pnet-eth		77.75 48	85.52 33	72.57 58	75.57 54	65.89 62	81.44 36	89.61 30	76.26 46

MVSCRF		77.21 49	74.57 45	78.98 50	79.29 50	83.87 45	75.01 44	74.12 54	73.78 52

MVE	76.82 20	76.63 50	77.52 41	76.03 55	70.40 60	85.49 43	75.68 41	79.36 44	72.20 54
Simon Fuhrmann, Fabian Langguth, Michael Goesele: MVE - A Multi-View Reconstruction Environment. EUROGRAPHICS Workshops on Graphics and Cultural Heritage (2014)
MVSNet_plusplus		76.50 51	66.38 54	83.25 39	91.25 29	73.79 57	53.62 63	79.14 45	84.70 33

example		74.48 52	65.59 55	80.41 48	84.10 48	83.16 46	62.82 51	68.36 60	73.98 51

Snet		74.39 53	64.28 58	81.13 43	88.04 34	73.18 60	60.15 56	68.41 59	82.18 36

MVSNet		73.65 54	70.01 50	76.07 54	76.39 53	77.59 53	61.93 53	78.09 47	74.23 50

MVSNet + Gipuma		72.61 55	71.20 48	73.54 56	75.39 55	73.46 59	65.60 48	76.81 52	71.78 55

firsttry		72.49 56	75.56 44	70.44 60	68.80 63	70.25 61	69.31 45	81.82 41	72.28 53

MVSNet_++		71.81 57	59.04 63	80.33 49	84.86 38	74.23 56	33.06 65	85.02 37	81.88 38

test_1124		71.63 58	63.97 59	76.73 52	69.39 62	85.55 42	56.73 58	71.21 56	75.26 47

CasMVSNet(SR_B)		70.07 59	71.18 49	69.33 61	69.73 61	77.06 54	64.06 50	78.31 46	61.20 62

F/T MVSNet+Gipuma		69.80 60	64.46 57	73.36 57	75.05 56	73.57 58	60.74 54	68.18 61	71.46 56

unMVSmet		66.94 61	60.67 62	71.11 59	71.91 59	79.70 51	54.62 60	66.73 62	61.74 61

confMetMVS		63.51 62	60.95 61	65.21 63	60.95 64	76.48 55	55.53 59	66.38 63	58.21 63

test_1120		58.74 63	57.03 64	59.88 64	49.42 66	64.79 63	49.65 64	64.41 64	65.45 60

PMVS	70.75 21	58.19 64	43.22 65	68.18 62	73.06 58	84.09 44	54.23 61	32.21 65	47.37 64
Y. Furukawa, J. Ponce: Accurate, dense, and robust multiview stereopsis. PAMI (2010)
Cas-MVS_preliminary		56.71 65	62.57 60	52.80 65	50.07 65	61.96 65	54.15 62	71.00 57	46.38 65

FADENet		18.58 66	15.51 66	20.63 66	21.45 67	32.74 67	15.09 66	15.93 66	7.68 66

CMPMVS	69.68 22	11.01 67	0.97 67	17.70 67	10.09 68	43.01 66	1.93 67	0.00 67	0.00 67
M. Jancosek, T. Pajdla: Multi-View Reconstruction Preserving Weakly-Supported Surfaces. CVPR 2011
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

UnsupFinetunedMVSNet					75.05 56

Low-resolution many-view benchmark