Optical flow evaluation results Statistics:     Average   SD   R2.5   R5.0   R10.0   A50   A75   A95  
Error type:   endpoint   angle   interpolation   normalized interpolation  
Show images: below table   above table        
R2.5
angle
error
avg. Army
(Hidden texture)
GT   im0   im1
Mequon
(Hidden texture)
GT   im0   im1
Schefflera
(Hidden texture)
GT   im0   im1
Wooden
(Hidden texture)
GT   im0   im1
Grove
(Synthetic)
GT   im0   im1
Urban
(Synthetic)
GT   im0   im1
Yosemite
(Synthetic)
GT   im0   im1
Teddy
(Stereo)
GT   im0   im1
rank all disc untext all disc untext all disc untext all disc untext all disc untext all disc untext all disc untext all disc untext
NNF-Local [87]13.5 20.0 3 43.3 3 13.0 3 15.9 10 45.3 25 16.1 14 12.5 4 34.1 9 12.8 9 9.68 19 32.0 11 7.01 23 23.3 4 30.0 4 16.2 4 22.0 31 45.4 9 23.8 60 25.8 18 44.8 6 17.6 17 4.37 9 14.4 27 0.48 1
NN-field [71]14.0 22.1 7 44.0 5 14.6 4 18.4 23 47.3 37 19.2 27 12.5 4 32.9 6 14.0 14 6.57 2 28.2 4 3.57 2 23.4 5 30.1 5 15.9 3 17.9 14 36.8 1 15.8 4 35.9 48 51.5 32 27.4 46 4.58 12 15.3 30 0.55 2
TC/T-Flow [76]18.8 19.9 2 46.9 10 10.2 1 16.0 11 47.8 39 13.9 5 13.0 8 36.6 20 11.1 4 8.90 12 35.0 34 6.10 12 27.2 17 35.7 22 21.0 14 15.7 4 47.2 17 15.8 4 21.0 6 39.2 2 42.1 73 7.86 40 19.5 43 10.9 52
ALD-Flow [66]19.2 21.6 6 46.0 7 15.9 6 15.5 8 41.3 9 15.6 10 13.1 9 35.2 12 12.2 5 8.17 10 33.4 20 5.35 9 28.1 24 37.3 33 20.3 10 16.4 6 47.4 19 16.0 6 26.5 21 45.4 8 41.8 71 8.39 44 22.3 53 11.3 54
ComponentFusion [96]19.4 20.0 3 46.3 8 14.8 5 16.6 14 40.3 6 18.5 24 11.7 3 33.6 7 10.9 3 7.09 5 35.2 36 4.45 5 27.6 20 36.0 25 21.4 17 21.5 27 54.1 51 20.2 29 31.8 41 56.8 58 16.2 10 5.46 22 12.8 20 7.47 27
nLayers [57]20.5 22.7 9 40.3 2 18.4 9 27.2 86 45.9 29 30.4 86 15.7 22 35.4 14 21.4 59 8.12 9 26.6 2 6.21 13 22.5 2 29.0 3 15.5 2 19.9 19 40.8 4 17.8 9 31.3 37 52.6 41 16.9 15 4.26 6 11.2 7 5.84 6
OFLAF [77]21.0 29.6 34 47.4 12 24.8 18 17.8 19 40.9 7 18.3 22 11.6 2 26.9 2 13.2 11 11.5 39 29.3 8 8.97 57 23.7 6 30.9 6 16.5 5 22.2 35 41.5 5 19.3 19 30.2 32 50.6 27 32.9 52 5.94 27 13.8 22 8.44 36
WLIF-Flow [93]21.3 27.0 19 47.0 11 23.1 16 21.7 43 46.6 35 23.3 41 14.8 13 36.2 16 16.3 22 9.33 16 32.2 13 6.64 17 27.7 21 35.1 20 23.4 29 21.6 28 47.7 21 19.2 16 28.7 28 45.3 7 32.0 51 4.50 11 11.2 7 6.43 11
RNLOD-Flow [121]21.3 20.7 5 43.8 4 19.4 10 18.1 20 45.9 29 17.0 19 12.7 6 34.1 9 12.3 6 7.38 6 28.7 6 4.87 6 26.5 13 34.9 18 20.9 12 20.3 20 46.2 12 20.2 29 43.1 72 60.3 68 47.5 90 4.98 19 12.7 19 6.55 14
HAST [109]21.7 19.5 1 40.1 1 11.6 2 16.1 12 39.7 4 14.8 8 8.49 1 21.2 1 7.09 1 6.79 3 29.0 7 3.66 3 21.6 1 28.3 2 13.8 1 24.3 54 48.5 25 24.3 62 41.7 69 59.8 67 63.0 114 6.05 29 11.5 13 8.72 39
MDP-Flow2 [68]23.5 35.3 46 55.4 41 30.2 41 14.2 3 39.7 4 14.2 7 13.6 10 31.4 3 12.9 10 12.2 44 34.1 28 8.19 45 27.7 21 35.0 19 22.1 20 22.4 39 45.4 9 21.5 42 27.1 22 54.1 45 16.5 11 5.87 26 14.0 24 4.25 3
OAR-Flow [125]24.4 25.6 12 54.9 38 22.4 14 18.7 24 44.8 22 19.1 25 17.2 35 43.7 44 18.0 28 8.53 11 31.6 10 5.65 10 29.7 34 39.3 38 20.9 12 14.5 2 47.3 18 13.4 2 14.1 2 38.0 1 20.7 30 9.84 57 21.1 49 16.1 68
Layers++ [37]24.6 28.0 31 48.8 15 30.9 44 23.3 58 45.6 28 25.5 65 13.7 11 31.4 3 18.1 29 8.08 8 24.9 1 5.87 11 22.9 3 28.1 1 19.6 8 22.2 35 46.3 14 20.7 33 39.6 64 55.9 54 35.0 54 4.16 3 9.78 2 6.81 16
TC-Flow [46]24.9 24.4 11 52.2 31 22.6 15 11.8 2 38.8 3 11.2 1 12.7 6 35.7 15 9.84 2 9.88 21 34.9 32 7.49 31 28.9 31 38.6 37 21.2 16 20.6 22 52.1 36 21.6 43 22.6 13 47.1 12 36.3 57 9.05 50 21.5 51 12.4 59
AGIF+OF [85]27.1 26.3 13 48.8 15 24.1 17 24.9 67 52.0 67 26.2 67 16.0 27 39.4 30 19.7 46 8.96 13 32.0 11 6.64 17 26.7 15 33.7 12 21.5 19 21.4 26 49.4 28 18.5 13 28.2 25 49.6 21 30.5 50 4.63 15 11.3 12 7.30 24
LME [70]27.1 31.5 36 51.4 26 21.0 13 14.7 5 36.9 2 15.7 11 16.0 27 35.2 12 19.8 49 11.8 41 36.6 40 8.08 40 28.6 28 36.0 25 25.5 40 21.1 25 49.0 27 19.8 26 29.8 31 50.0 24 21.5 32 6.61 31 15.1 28 7.95 32
PH-Flow [101]28.0 26.7 15 51.6 27 25.6 24 21.7 43 49.4 49 23.8 46 15.5 17 37.6 23 19.5 42 10.2 26 33.8 24 7.44 30 26.4 11 33.7 12 21.0 14 22.0 31 50.3 30 20.4 31 38.8 58 48.4 17 45.4 80 4.26 6 11.2 7 6.38 10
FC-2Layers-FF [74]28.1 26.9 18 48.6 14 28.3 35 22.1 47 48.8 44 23.8 46 14.1 12 32.4 5 19.6 44 9.10 14 28.3 5 6.47 16 25.5 7 31.7 7 23.0 27 23.3 44 47.7 21 21.8 44 44.1 75 56.3 57 46.0 85 3.54 1 9.08 1 5.44 5
Classic+CPF [83]29.2 27.1 23 51.2 25 25.5 23 23.9 62 51.4 63 25.1 60 15.9 24 39.3 29 19.4 40 9.17 15 32.8 16 6.69 19 27.3 18 34.5 16 23.9 30 21.0 24 48.7 26 18.0 11 35.6 47 49.9 23 45.7 82 4.25 5 10.7 4 6.61 15
NNF-EAC [103]29.4 34.9 41 54.8 36 29.9 40 15.6 9 41.7 11 15.9 12 15.1 16 34.8 11 15.7 16 12.4 48 35.1 35 8.33 48 28.1 24 35.7 22 22.9 25 24.5 57 46.2 12 22.7 54 31.6 40 49.1 19 20.1 28 7.54 38 17.1 38 7.44 26
COFM [59]29.5 22.1 7 49.1 17 16.6 7 18.2 21 43.5 16 19.2 27 15.9 24 38.3 26 21.5 60 7.02 4 32.6 14 4.40 4 31.4 42 37.9 34 35.0 76 22.1 33 46.4 15 18.3 12 28.7 28 45.9 10 45.5 81 9.25 51 15.5 32 15.7 67
Sparse-NonSparse [56]30.0 26.8 17 51.9 30 26.8 30 22.2 52 49.0 47 24.6 56 15.6 19 39.4 30 19.0 35 9.46 18 33.6 22 7.07 25 29.0 32 37.2 32 24.3 35 21.7 30 50.4 31 19.5 24 34.0 43 45.8 9 41.6 69 4.44 10 10.9 5 6.96 20
IROF++ [58]30.1 27.4 25 50.7 21 26.7 29 22.1 47 50.1 58 24.1 50 16.3 30 40.0 34 19.6 44 10.6 33 34.3 29 7.72 36 27.9 23 35.7 22 22.5 21 22.3 37 54.1 51 19.9 27 25.5 17 49.6 21 11.5 6 5.49 23 14.1 25 6.54 13
FESL [72]30.8 27.0 19 46.3 8 31.2 46 26.0 77 51.8 66 26.9 71 15.8 23 37.0 21 19.5 42 7.89 7 30.7 9 5.17 8 26.6 14 33.8 14 22.6 22 20.3 20 45.8 11 19.3 19 39.9 65 61.2 72 35.6 55 5.04 20 12.5 18 6.48 12
Efficient-NL [60]31.5 23.3 10 44.7 6 17.6 8 24.6 66 51.6 65 25.4 64 15.0 15 36.2 16 17.5 26 9.92 22 33.1 18 6.94 21 26.4 11 34.0 15 20.3 10 27.2 66 49.4 28 22.6 49 37.6 53 50.5 26 37.1 58 6.98 36 16.2 34 8.16 33
LSM [39]33.1 26.5 14 50.8 22 27.0 31 22.1 47 49.4 49 24.4 54 15.6 19 38.2 25 19.4 40 10.1 23 33.0 17 7.43 29 28.6 28 36.3 27 24.7 36 22.8 42 50.5 32 20.9 37 40.3 67 49.3 20 45.8 84 4.76 17 12.0 15 6.93 19
PMMST [114]33.5 42.4 69 58.9 45 40.2 65 22.1 47 45.3 25 24.7 59 17.9 41 38.4 27 18.8 32 12.2 44 28.1 3 8.26 46 25.7 8 32.7 10 18.5 7 22.3 37 45.3 8 20.8 35 28.5 27 50.7 28 18.3 19 8.51 47 16.8 35 8.73 40
Classic+NL [31]34.4 27.2 24 47.8 13 28.4 36 22.0 46 49.4 49 24.1 50 15.5 17 37.4 22 19.8 49 10.4 30 33.8 24 7.40 28 28.5 26 36.4 28 24.2 34 23.3 44 52.2 37 21.1 38 43.9 73 52.5 38 44.6 77 4.60 14 11.2 7 7.04 22
Ramp [62]34.5 27.0 19 52.3 32 26.1 27 22.3 54 49.1 48 24.6 56 15.6 19 37.8 24 19.8 49 10.5 32 33.6 22 7.61 33 28.6 28 36.6 29 24.0 31 23.6 47 51.2 33 21.8 44 38.2 54 44.5 4 46.1 86 4.86 18 12.1 16 7.13 23
2DHMM-SAS [92]37.3 26.7 15 51.8 29 25.6 24 22.2 52 51.3 61 23.8 46 17.8 39 42.3 39 20.0 54 10.4 30 34.6 31 7.52 32 28.5 26 36.6 29 24.0 31 22.7 41 54.5 54 20.6 32 39.0 60 47.8 15 45.1 78 5.51 24 13.9 23 7.73 30
FMOF [94]37.5 27.8 30 50.4 20 27.1 33 26.2 79 52.5 71 27.4 74 15.9 24 36.4 19 21.7 61 9.71 20 32.7 15 6.98 22 27.3 18 34.6 17 24.1 33 23.7 49 47.6 20 19.7 25 38.4 55 55.4 51 48.9 93 5.80 25 14.1 25 7.00 21
SVFilterOh [111]38.6 39.6 59 54.6 35 39.8 64 23.8 61 44.7 21 24.0 49 17.3 36 33.8 8 18.8 32 10.2 26 36.4 39 4.93 7 25.9 10 32.1 8 19.6 8 24.7 58 46.7 16 22.9 57 52.0 97 76.3 115 59.2 108 4.23 4 10.9 5 5.01 4
S2D-Matching [84]39.6 27.7 27 50.2 19 28.4 36 22.1 47 48.8 44 24.3 52 16.5 32 40.0 34 19.3 38 10.6 33 33.8 24 7.78 37 29.1 33 36.7 31 25.0 37 24.4 56 53.0 47 22.6 49 47.0 86 53.6 43 50.8 98 4.58 12 11.2 7 7.54 28
ProbFlowFields [128]40.0 33.9 37 68.5 67 30.9 44 20.3 33 45.3 25 22.1 37 19.2 48 44.8 47 22.9 65 11.7 40 39.3 48 8.91 56 31.0 40 40.1 42 23.3 28 16.8 7 47.8 23 19.2 16 23.7 16 55.9 54 23.9 40 8.39 44 22.4 54 9.82 48
SimpleFlow [49]40.4 28.6 33 51.6 27 29.5 39 25.0 68 51.3 61 28.2 78 18.6 46 43.0 42 23.3 66 10.1 23 33.5 21 7.04 24 29.9 35 37.9 34 26.2 42 27.8 71 52.2 37 24.0 61 35.1 46 47.9 16 29.7 49 4.66 16 12.4 17 6.92 18
PMF [73]41.3 37.7 54 58.3 43 27.3 34 19.4 26 45.0 24 18.3 22 16.2 29 39.9 33 14.0 14 13.6 59 35.7 37 8.03 39 25.8 9 32.5 9 17.1 6 30.2 78 57.0 63 31.6 80 58.9 105 74.7 108 55.9 105 3.95 2 10.3 3 6.15 9
Adaptive [20]42.2 27.0 19 52.6 33 19.8 11 21.9 45 47.8 39 22.6 39 20.5 53 47.8 55 19.7 46 10.3 28 39.9 51 6.31 15 45.6 110 52.6 108 51.3 109 17.4 8 48.2 24 13.9 3 34.8 45 56.9 60 19.8 26 6.04 28 15.2 29 7.54 28
TV-L1-MCT [64]42.8 27.5 26 49.4 18 27.0 31 26.5 80 52.8 76 27.8 77 16.8 33 39.1 28 21.8 62 10.6 33 33.8 24 7.82 38 30.3 37 38.1 36 28.7 56 24.7 58 53.4 49 23.4 58 27.4 23 52.5 38 19.4 23 7.28 37 15.3 30 11.4 56
Correlation Flow [75]42.9 36.6 51 55.3 40 34.4 54 16.6 14 44.4 20 14.8 8 18.3 45 42.9 41 12.7 8 12.4 48 39.7 50 8.46 52 32.2 45 40.6 45 25.2 38 29.0 74 54.2 53 29.7 77 39.1 61 52.1 36 47.4 89 6.53 30 16.1 33 6.88 17
IROF-TV [53]43.5 30.9 35 54.8 36 31.6 48 22.8 57 50.5 59 25.1 60 16.9 34 40.8 36 20.8 56 14.0 61 43.7 65 10.1 61 31.2 41 39.5 39 28.6 54 26.8 65 58.7 72 25.6 65 18.8 4 48.4 17 8.08 5 5.28 21 13.6 21 7.77 31
AggregFlow [97]43.6 36.3 50 52.7 34 35.7 56 26.6 81 52.6 74 26.7 69 23.3 66 48.2 57 28.9 82 12.1 42 34.9 32 8.63 54 30.2 36 40.3 44 21.4 17 15.9 5 38.3 2 16.8 7 26.1 20 47.3 13 16.8 13 12.6 69 20.3 45 20.0 78
Occlusion-TV-L1 [63]44.2 34.3 40 58.5 44 25.1 19 20.0 29 46.5 34 20.8 33 22.3 63 49.9 58 20.6 55 12.6 54 41.7 57 8.41 51 35.2 63 44.9 72 32.3 67 17.7 10 52.6 42 21.1 38 28.2 25 52.5 38 13.0 7 9.66 53 23.7 58 10.6 50
Classic++ [32]47.5 27.7 27 51.0 24 28.7 38 21.5 42 45.9 29 24.3 52 18.1 44 44.3 45 19.9 52 10.3 28 37.7 42 7.14 26 33.4 48 44.1 63 27.9 50 24.0 52 57.9 68 21.4 41 46.3 82 55.6 52 49.7 95 8.45 46 20.7 46 9.69 47
Aniso-Texture [82]47.5 28.5 32 50.9 23 32.7 51 21.2 37 41.7 11 25.3 63 17.6 38 41.2 37 21.1 58 5.72 1 33.2 19 2.54 1 35.1 62 43.7 60 31.3 65 23.6 47 53.6 50 22.6 49 62.2 112 75.8 113 53.5 102 6.89 35 17.4 39 8.20 34
MDP-Flow [26]48.5 35.5 47 65.0 54 32.4 49 20.6 35 43.8 18 24.4 54 18.0 43 43.4 43 19.9 52 14.9 69 41.8 59 11.5 70 30.8 39 39.6 40 25.3 39 23.8 50 57.4 66 22.2 46 31.0 35 59.3 65 16.8 13 10.8 64 26.5 63 10.9 52
DeepFlow2 [108]49.0 39.1 57 66.4 58 44.4 73 20.1 31 47.9 41 20.9 34 23.8 69 52.8 65 26.3 75 12.2 44 43.2 64 7.64 34 31.4 42 41.7 52 22.9 25 18.2 16 52.4 39 17.9 10 29.6 30 44.7 5 39.0 62 16.2 83 31.2 83 22.7 84
OFH [38]49.1 41.9 68 61.6 49 48.5 79 14.7 5 42.7 14 14.1 6 17.4 37 47.6 53 12.6 7 10.6 33 38.5 44 8.39 50 34.8 60 43.7 60 34.5 74 27.2 66 61.9 85 29.5 76 21.3 9 57.6 61 21.4 31 12.2 68 29.8 75 16.2 69
BriefMatch [124]50.5 34.1 39 59.4 46 30.2 41 17.1 16 43.6 17 16.1 14 14.8 13 36.3 18 13.4 12 9.41 17 34.3 29 6.26 14 33.4 48 41.2 49 31.8 66 40.3 105 63.5 87 42.7 108 47.2 87 61.1 70 59.1 107 12.7 71 23.4 55 21.6 82
CPM-Flow [116]51.2 35.2 42 67.0 60 25.3 20 25.8 74 53.6 80 28.2 78 25.5 72 58.5 85 27.5 76 12.4 48 48.0 76 8.11 43 33.9 53 44.2 64 27.1 44 17.6 9 51.8 34 19.1 14 21.3 9 51.7 33 22.6 36 9.92 58 27.2 67 11.3 54
PGM-C [120]51.8 35.2 42 67.1 62 25.4 22 25.8 74 53.6 80 28.2 78 25.8 75 59.3 88 27.5 76 12.4 48 48.1 77 8.16 44 33.9 53 44.3 66 27.1 44 17.7 10 52.8 46 19.1 14 20.8 5 50.3 25 22.4 34 9.82 56 27.2 67 11.5 58
S2F-IF [123]52.0 35.7 48 69.0 70 26.6 28 24.1 64 54.0 87 26.0 66 25.6 74 60.9 90 25.9 71 12.4 48 46.6 73 8.38 49 34.3 56 44.2 64 27.7 49 18.0 15 53.2 48 19.2 16 21.2 8 51.7 33 22.9 37 8.97 49 24.9 61 9.11 43
CostFilter [40]53.6 43.6 75 65.7 56 40.8 66 20.8 36 45.9 29 21.0 35 18.8 47 44.9 48 18.3 31 17.3 78 39.5 49 13.9 80 26.7 15 32.8 11 22.7 24 30.9 81 60.0 79 31.6 80 59.7 106 81.5 122 59.3 109 4.31 8 11.9 14 5.93 7
RFlow [90]53.6 44.4 77 75.2 90 50.5 83 16.5 13 42.1 13 17.2 20 21.4 57 52.2 61 16.0 17 11.3 38 36.2 38 7.25 27 35.2 63 44.3 66 32.3 67 24.3 54 55.6 58 22.6 49 38.6 56 55.3 49 41.6 69 13.4 73 29.9 76 16.9 73
EpicFlow [102]53.7 35.2 42 67.2 63 25.3 20 25.8 74 53.8 83 28.2 78 26.1 77 60.1 89 27.5 76 12.4 48 48.1 77 8.10 42 34.2 55 44.5 68 28.0 51 17.8 12 52.7 44 19.4 21 21.0 6 52.1 36 22.5 35 10.4 61 27.3 69 12.7 61
TV-L1-improved [17]55.3 27.7 27 57.9 42 20.5 12 18.2 21 44.9 23 19.1 25 19.4 49 47.7 54 17.0 25 10.1 23 38.5 44 6.75 20 35.9 70 46.0 78 27.3 46 43.7 111 70.2 108 47.5 112 51.4 94 60.5 69 50.3 97 10.3 60 26.8 66 10.8 51
FlowFields [110]55.9 35.7 48 68.7 68 25.8 26 25.6 73 55.1 92 27.7 76 26.8 80 62.8 93 27.6 79 13.0 55 46.8 74 9.13 58 34.6 59 44.9 72 28.1 52 17.8 12 54.9 56 19.4 21 21.3 9 54.8 47 24.2 41 9.25 51 26.4 62 8.47 37
Steered-L1 [118]57.8 38.7 56 67.9 64 43.1 71 11.6 1 34.7 1 12.3 2 16.3 30 41.3 38 13.9 13 12.1 42 38.6 46 8.30 47 34.5 58 43.4 57 32.5 69 29.4 76 61.1 84 25.9 66 60.6 109 67.0 90 70.1 120 15.9 82 30.6 79 24.0 86
Sparse Occlusion [54]58.4 38.6 55 61.8 50 32.5 50 26.0 77 48.8 44 29.5 85 20.0 51 45.2 49 19.2 37 14.3 65 38.4 43 9.67 60 34.4 57 42.6 54 26.7 43 25.4 61 52.4 39 22.4 48 67.3 117 75.9 114 48.3 91 8.07 42 19.9 44 7.36 25
DeepFlow [86]59.6 47.3 83 71.9 80 64.0 97 21.4 40 48.2 42 22.7 40 27.9 84 58.2 82 31.6 86 15.1 70 42.7 60 10.6 66 31.5 44 42.1 53 22.6 22 19.6 18 56.7 62 19.4 21 27.6 24 46.2 11 39.7 63 20.6 91 35.3 98 28.0 94
MLDP_OF [89]60.0 48.8 86 77.3 93 52.2 85 20.1 31 49.6 54 19.3 29 23.5 68 54.6 69 18.9 34 12.3 47 38.6 46 7.65 35 33.5 51 40.9 46 29.1 60 28.4 72 55.9 59 31.6 80 49.1 91 62.2 76 60.5 111 7.91 41 16.8 35 8.95 41
TF+OM [100]60.9 39.8 60 55.2 39 30.4 43 20.4 34 41.0 8 23.5 43 19.5 50 39.4 30 28.0 80 18.4 81 37.0 41 18.0 86 35.0 61 41.1 47 39.6 88 29.6 77 52.6 42 29.1 75 49.0 90 66.8 89 43.6 76 14.4 75 29.3 72 18.0 75
CombBMOF [113]62.9 42.4 69 71.5 78 31.3 47 25.2 69 52.9 77 25.1 60 17.9 41 45.8 50 16.1 18 14.2 63 41.7 57 11.4 69 33.5 51 40.2 43 29.6 61 34.5 88 59.9 78 37.3 94 55.2 103 69.9 99 46.5 88 6.78 33 16.8 35 8.62 38
EPPM w/o HM [88]63.4 43.1 73 72.4 82 38.1 62 18.7 24 52.5 71 16.9 18 21.3 56 56.2 74 17.5 26 16.2 74 45.3 71 12.7 77 33.4 48 39.6 40 30.0 62 33.5 84 65.4 93 33.8 86 45.5 77 66.1 87 65.5 117 6.88 34 18.1 40 9.16 45
Complementary OF [21]64.5 51.9 93 74.9 89 59.3 93 14.2 3 41.6 10 13.7 3 20.4 52 46.6 51 19.7 46 22.2 88 40.8 53 21.0 91 36.0 72 43.4 57 38.5 86 33.9 87 63.8 88 31.6 80 31.1 36 51.9 35 36.2 56 18.9 88 34.4 96 29.4 96
Aniso. Huber-L1 [22]65.4 33.9 37 65.1 55 32.8 52 34.0 92 54.0 87 40.0 92 27.9 84 55.0 71 38.4 91 15.2 71 49.9 81 12.0 72 35.3 65 44.6 69 28.5 53 23.9 51 55.9 59 20.7 33 50.6 92 62.1 75 39.7 63 8.15 43 20.7 46 8.39 35
ComplOF-FED-GPU [35]67.4 49.5 88 75.7 92 55.3 88 15.3 7 47.0 36 13.8 4 21.1 55 52.7 64 16.1 18 17.1 77 40.6 52 14.2 81 35.7 69 45.1 75 32.5 69 35.3 91 67.5 101 34.4 88 46.5 84 59.0 64 50.8 98 12.8 72 29.6 73 16.6 72
ACK-Prior [27]67.6 55.9 95 72.7 84 59.3 93 17.5 18 43.3 15 16.0 13 17.8 39 42.3 39 16.3 22 17.6 79 41.3 55 12.0 72 35.3 65 41.1 47 35.9 77 37.4 101 59.6 76 34.3 87 59.8 108 61.1 70 74.7 121 17.7 87 29.2 71 27.4 89
TCOF [69]67.6 45.0 78 70.0 72 51.5 84 25.5 72 53.7 82 26.7 69 26.7 78 56.2 74 32.0 87 21.9 87 43.1 63 22.2 93 37.8 82 48.9 94 25.7 41 18.8 17 44.7 7 20.0 28 52.1 98 67.5 92 25.8 42 10.7 63 26.7 64 11.4 56
Rannacher [23]67.7 43.1 73 71.0 76 45.2 75 24.1 64 49.4 49 26.4 68 26.0 76 56.9 78 26.0 73 14.2 63 42.7 60 10.5 65 37.1 79 47.9 89 30.7 64 32.3 82 65.2 92 27.0 70 44.0 74 56.0 56 39.7 63 7.83 39 21.2 50 9.19 46
F-TV-L1 [15]69.9 66.8 103 84.2 102 77.3 105 27.1 84 52.0 67 29.1 84 27.2 81 57.3 80 24.2 69 24.1 92 52.0 85 19.5 89 39.3 89 47.7 88 39.3 87 24.2 53 56.5 61 24.9 63 33.2 42 53.7 44 20.1 28 6.69 32 18.6 41 5.94 8
ROF-ND [107]70.0 49.2 87 71.5 78 49.4 81 22.5 56 46.2 33 20.6 31 21.4 57 50.0 59 18.1 29 23.1 90 53.7 89 16.0 83 35.9 70 46.1 80 28.6 54 33.5 84 58.5 70 30.3 78 60.7 110 70.5 100 60.4 110 9.67 54 20.9 48 10.2 49
LDOF [28]70.0 41.7 67 70.7 73 47.2 78 24.0 63 53.9 86 24.6 56 26.7 78 58.4 84 25.5 70 15.8 72 57.4 95 10.2 63 36.0 72 45.3 76 34.8 75 22.1 33 58.1 69 21.3 40 30.6 33 56.8 58 23.4 38 22.5 100 38.6 105 30.2 97
SIOF [67]71.0 50.3 91 66.0 57 47.1 77 19.8 28 48.4 43 20.7 32 29.8 91 55.1 72 32.6 89 25.9 94 48.3 79 25.0 95 37.7 80 46.4 82 36.8 80 32.9 83 58.6 71 35.6 92 37.2 52 53.1 42 18.6 20 16.8 85 33.0 88 21.3 81
LocallyOriented [52]71.5 39.4 58 60.5 47 35.7 56 27.9 87 57.8 98 28.5 82 28.1 87 58.3 83 30.6 85 13.9 60 41.3 55 10.1 61 37.8 82 47.3 86 33.0 73 24.8 60 52.5 41 28.1 72 39.4 63 62.4 77 37.5 59 15.7 80 33.0 88 18.2 77
Second-order prior [8]72.2 37.6 53 70.9 75 37.2 60 22.4 55 51.2 60 23.7 45 24.5 71 59.1 87 22.6 63 11.2 37 41.2 54 8.48 53 37.9 84 49.6 101 28.8 57 29.0 74 68.8 105 26.0 67 55.5 104 64.7 82 52.7 101 14.7 77 34.1 94 17.3 74
CRTflow [80]72.2 40.9 62 72.5 83 36.2 58 21.3 39 49.4 49 21.8 36 22.9 64 57.5 81 19.0 35 14.0 61 44.7 68 10.3 64 35.4 68 45.0 74 30.4 63 46.6 115 73.4 110 53.8 116 38.8 58 65.5 84 38.2 60 19.5 90 38.5 104 27.6 92
Brox et al. [5]72.5 43.7 76 74.4 87 56.4 90 27.0 83 52.5 71 30.5 87 23.4 67 54.5 68 23.3 66 13.4 56 50.0 82 8.66 55 39.8 92 46.5 83 47.6 104 21.6 28 59.8 77 22.8 56 30.9 34 59.3 65 7.61 3 23.1 103 37.0 102 33.4 104
NL-TV-NCC [25]72.5 46.1 80 68.3 66 43.4 72 25.2 69 55.3 93 23.5 43 21.8 62 47.2 52 16.9 24 16.9 76 44.1 66 11.9 71 38.5 86 49.1 98 27.3 46 36.5 97 65.7 94 34.9 90 46.2 79 75.7 112 45.7 82 12.6 69 28.8 70 9.10 42
FlowNetS+ft+v [112]72.5 36.8 52 67.0 60 39.4 63 25.3 71 52.1 69 27.6 75 27.8 83 57.2 79 35.5 90 13.5 57 50.9 83 9.44 59 40.1 93 49.5 100 36.8 80 20.9 23 57.0 63 20.8 35 46.3 82 66.3 88 41.0 67 17.4 86 34.3 95 24.1 87
DF-Auto [115]72.6 41.6 66 64.2 53 32.8 52 42.0 97 58.8 101 49.6 98 34.8 97 62.2 92 47.2 98 20.7 85 53.4 88 14.5 82 36.2 75 44.7 70 37.2 83 15.1 3 42.9 6 17.4 8 45.3 76 69.0 96 13.0 7 24.2 106 37.7 103 31.9 100
TriangleFlow [30]73.3 41.5 65 63.2 51 42.6 70 21.4 40 52.4 70 20.2 30 21.7 61 53.6 67 16.2 20 14.8 67 44.4 67 10.9 67 43.2 103 52.9 110 43.4 95 36.8 100 65.9 95 38.8 97 42.2 70 65.4 83 41.8 71 15.8 81 35.2 97 22.5 83
SRR-TVOF-NL [91]73.5 47.6 84 69.1 71 41.9 69 23.3 58 52.7 75 23.3 41 25.5 72 56.8 77 25.9 71 13.5 57 47.9 75 8.09 41 36.9 78 43.8 62 32.9 72 25.8 62 57.7 67 22.6 49 62.8 114 75.2 110 49.1 94 21.0 93 29.7 74 31.6 99
DPOF [18]74.3 45.3 79 68.9 69 37.6 61 26.7 82 56.9 95 26.9 71 24.3 70 54.6 69 26.2 74 18.6 82 54.6 92 13.6 79 33.3 47 43.1 55 28.9 58 27.6 70 60.8 81 26.3 68 47.2 87 55.3 49 76.0 123 14.3 74 31.0 82 15.3 66
Bartels [41]75.3 48.6 85 63.2 51 61.4 96 23.4 60 44.0 19 27.1 73 21.4 57 44.6 46 23.8 68 26.2 95 43.0 62 25.4 96 36.8 76 44.7 70 41.7 92 33.7 86 60.0 79 41.7 105 52.6 99 67.2 91 61.1 112 11.2 65 23.4 55 16.3 70
Dynamic MRF [7]75.5 49.5 88 78.0 94 55.8 89 17.2 17 47.4 38 16.5 17 21.6 60 56.3 76 16.2 20 14.8 67 46.4 72 12.5 75 41.2 97 49.0 96 45.1 99 35.3 91 70.7 109 38.5 96 37.1 51 57.7 62 55.1 103 21.3 95 36.7 101 31.2 98
SuperFlow [81]75.6 35.2 42 61.1 48 34.4 54 35.2 93 53.5 79 42.5 93 27.7 82 52.5 63 43.5 97 27.5 96 60.5 99 27.6 97 36.0 72 43.3 56 42.2 93 22.6 40 59.2 75 22.2 46 46.2 79 62.0 74 23.4 38 21.8 98 36.0 99 32.5 102
CBF [12]76.2 41.4 63 74.0 86 48.5 79 40.2 96 51.5 64 51.7 100 22.9 64 50.8 60 28.5 81 14.3 65 44.7 68 11.2 68 38.3 85 46.1 80 36.1 78 26.3 64 55.1 57 24.9 63 61.5 111 71.0 101 52.0 100 11.6 67 26.7 64 14.8 65
CLG-TV [48]76.5 41.4 63 68.0 65 40.8 66 37.0 94 53.1 78 45.3 94 30.9 92 58.7 86 40.2 93 22.8 89 62.0 100 19.3 88 39.0 88 47.6 87 38.2 85 27.2 66 61.0 83 27.1 71 46.2 79 57.8 63 29.3 48 9.74 55 24.6 60 9.11 43
Local-TV-L1 [65]76.5 56.8 97 79.1 95 74.5 101 39.5 95 53.8 83 46.1 95 38.1 98 66.2 96 43.1 96 23.9 91 52.9 87 21.1 92 32.3 46 41.4 51 27.5 48 23.2 43 54.8 55 22.7 54 25.8 18 47.5 14 33.4 53 26.9 109 40.5 107 40.5 113
CNN-flow-warp+ref [117]76.8 42.6 71 71.2 77 49.8 82 31.6 91 53.8 83 37.3 90 32.7 94 63.8 94 42.7 95 16.0 73 55.1 93 12.1 74 38.6 87 46.0 78 43.8 97 23.3 44 59.1 74 23.6 59 23.5 15 50.9 30 21.9 33 24.2 106 36.2 100 32.9 103
HBM-GC [105]78.7 73.5 111 79.7 96 79.4 107 30.0 88 49.7 56 33.6 88 29.3 90 47.8 55 30.4 83 35.4 99 45.2 70 33.3 99 30.5 38 35.1 20 32.6 71 34.6 89 52.0 35 35.4 91 70.9 120 80.1 118 62.6 113 8.83 48 19.1 42 13.0 62
TriFlow [95]82.0 47.1 82 66.5 59 41.1 68 31.2 90 49.6 54 37.3 90 27.9 84 52.3 62 39.4 92 24.7 93 49.3 80 22.3 94 37.7 80 43.5 59 43.1 94 28.4 72 52.7 44 29.0 74 76.7 124 73.7 105 99.5 127 16.2 83 30.5 78 20.1 79
p-harmonic [29]83.9 50.4 92 86.6 112 56.5 92 27.1 84 54.5 89 28.9 83 32.9 95 69.3 102 30.4 83 19.8 83 65.1 103 16.2 84 39.6 91 47.2 85 40.3 90 30.4 79 66.5 97 32.2 84 45.6 78 64.4 80 28.9 47 10.2 59 24.1 59 13.2 63
Learning Flow [11]84.3 42.9 72 70.7 73 44.8 74 30.2 89 54.7 90 34.7 89 28.2 88 55.9 73 32.3 88 17.6 79 57.1 94 12.6 76 44.0 106 52.6 108 47.8 106 30.5 80 64.6 90 30.3 78 46.9 85 63.3 78 42.8 75 14.7 77 31.6 86 16.3 70
Fusion [6]85.6 40.5 61 75.6 91 45.9 76 20.0 29 50.0 57 22.5 38 20.8 54 52.8 65 22.8 64 16.2 74 52.8 86 13.5 78 43.1 102 49.0 96 47.5 101 39.6 104 67.8 103 43.8 110 63.9 115 75.0 109 46.3 87 35.5 118 42.6 113 53.3 123
StereoFlow [44]85.9 95.9 127 96.0 126 97.4 127 88.3 127 96.2 127 86.2 123 82.6 124 94.8 126 73.7 121 91.4 126 96.3 126 90.3 126 53.0 117 61.6 121 52.8 110 11.2 1 39.3 3 11.7 1 10.5 1 42.5 3 1.70 1 11.5 66 23.6 57 18.0 75
Shiralkar [42]87.5 46.1 80 85.6 108 54.3 87 19.7 27 57.7 97 18.2 21 28.2 88 70.8 103 19.3 38 20.5 84 59.0 97 18.4 87 39.5 90 49.7 102 36.3 79 40.4 106 76.1 112 41.2 104 51.9 96 65.8 85 64.2 115 21.0 93 42.4 112 25.3 88
SegOF [10]90.6 56.3 96 71.9 80 37.1 59 57.3 111 62.9 108 68.3 113 46.0 104 69.0 101 57.2 110 41.0 101 59.5 98 37.2 100 43.5 104 48.3 92 56.4 114 38.2 102 69.6 107 39.1 98 17.9 3 64.5 81 3.40 2 22.7 102 33.0 88 32.0 101
StereoOF-V1MT [119]91.2 49.7 90 86.0 109 56.4 90 21.2 37 68.8 111 16.3 16 32.2 93 80.7 110 20.8 56 21.3 86 66.1 105 17.4 85 47.0 112 57.0 112 47.5 101 41.7 109 81.2 116 40.9 102 38.6 56 68.1 94 48.6 92 23.2 104 42.2 111 27.7 93
Ad-TV-NDC [36]91.7 73.7 112 85.4 106 89.5 122 56.9 110 60.4 104 67.5 112 51.0 108 75.9 106 57.6 111 45.7 104 65.7 104 47.9 107 35.3 65 45.3 76 28.9 58 27.3 69 57.2 65 28.2 73 34.6 44 55.0 48 27.3 45 34.0 115 48.7 119 47.5 117
Modified CLG [34]93.9 68.7 106 80.5 97 76.1 103 52.0 107 61.0 105 63.6 109 51.9 109 79.4 108 55.6 108 47.4 107 72.1 110 46.7 106 41.2 97 49.7 102 46.0 100 26.0 63 64.7 91 26.7 69 31.4 39 55.6 52 19.9 27 29.0 112 43.8 115 39.9 112
IAOF2 [51]95.2 54.9 94 73.7 85 53.9 86 42.6 98 58.3 99 50.7 99 33.9 96 61.9 91 42.1 94 64.4 116 75.7 113 74.3 118 41.5 99 49.9 105 37.1 82 36.4 96 64.0 89 34.4 88 59.7 106 69.6 98 41.3 68 19.4 89 33.4 92 23.0 85
Filter Flow [19]97.4 62.9 98 74.4 87 60.8 95 42.8 99 60.1 103 49.4 97 42.5 100 66.0 95 51.2 101 52.1 112 69.5 107 50.2 108 44.8 108 49.7 102 54.4 112 41.9 110 66.7 98 43.6 109 74.3 123 88.9 125 42.6 74 10.6 62 21.6 52 12.6 60
GroupFlow [9]98.0 66.4 101 85.2 105 80.8 110 61.6 114 75.4 118 69.0 114 51.9 109 83.6 113 57.0 109 33.5 98 63.9 101 32.5 98 49.6 113 61.0 116 39.9 89 51.3 120 81.7 117 59.4 120 22.8 14 51.1 31 16.5 11 28.0 111 41.9 110 37.9 110
SPSA-learn [13]98.9 65.1 99 87.4 115 72.7 100 45.7 103 59.3 102 53.4 103 45.2 103 74.7 105 52.2 103 41.6 102 69.9 109 42.5 102 42.7 101 48.9 94 48.7 108 38.8 103 69.0 106 42.5 107 39.1 61 61.9 73 19.3 22 36.0 119 45.6 116 48.3 118
2D-CLG [1]99.0 77.2 115 82.3 100 75.4 102 61.5 113 65.9 110 73.7 117 63.2 118 89.6 120 60.8 116 82.8 124 88.3 121 86.8 124 43.5 104 49.3 99 54.8 113 35.1 90 67.5 101 36.1 93 21.3 9 50.8 29 15.5 9 34.4 117 46.3 118 46.0 115
IAOF [50]99.5 66.3 100 81.3 99 77.8 106 50.1 106 58.4 100 59.7 107 45.0 102 74.1 104 49.6 100 50.8 109 68.2 106 58.0 114 40.2 94 48.7 93 37.8 84 36.7 98 66.9 99 33.5 85 54.9 101 63.5 79 40.8 66 30.1 114 41.3 109 43.5 114
BlockOverlap [61]99.6 77.2 115 86.0 109 82.8 112 48.2 104 55.8 94 57.6 106 46.9 106 66.9 97 51.9 102 49.1 108 54.1 90 51.2 109 36.8 76 41.3 50 47.5 101 40.5 107 59.0 73 39.6 100 68.9 119 80.2 119 65.1 116 20.8 92 30.8 80 34.9 106
HBpMotionGpu [43]100.1 67.0 104 80.7 98 72.3 99 55.3 109 57.3 96 66.7 111 44.7 101 67.2 99 54.1 106 39.5 100 57.8 96 38.4 101 42.0 100 48.2 91 48.3 107 35.9 93 60.9 82 39.2 99 65.1 116 72.0 103 50.1 96 22.6 101 32.5 87 36.9 107
Black & Anandan [4]101.8 70.3 108 88.0 116 84.1 113 45.5 102 61.4 106 52.0 101 47.4 107 77.3 107 52.9 104 42.3 103 77.5 114 42.8 103 44.0 106 51.8 106 45.0 98 35.9 93 75.9 111 38.3 95 50.8 93 71.3 102 17.8 18 29.8 113 42.7 114 37.6 109
GraphCuts [14]102.0 66.6 102 87.0 113 80.0 109 43.1 100 63.0 109 46.1 95 41.8 99 67.0 98 53.4 105 28.5 97 64.0 102 20.8 90 40.2 94 48.1 90 43.5 96 46.5 113 63.4 86 40.5 101 62.7 113 75.4 111 69.5 119 23.8 105 33.3 91 38.5 111
FlowNet2 [122]103.6 78.8 117 86.2 111 79.4 107 63.5 118 70.2 113 72.8 116 57.2 115 82.0 111 59.7 115 46.2 105 51.4 84 45.6 105 45.3 109 54.0 111 41.1 91 36.7 98 67.3 100 41.0 103 67.6 118 80.8 120 55.1 103 14.5 76 30.0 77 13.9 64
Nguyen [33]105.8 75.6 114 85.4 106 85.8 115 67.0 119 61.6 107 83.0 121 57.1 114 80.2 109 64.1 118 70.8 117 80.2 116 77.4 120 45.9 111 52.2 107 56.4 114 36.3 95 68.1 104 42.0 106 41.4 68 66.0 86 19.7 25 34.1 116 45.6 116 46.1 116
SILK [79]105.8 72.5 109 85.1 104 88.3 119 61.9 115 71.8 114 73.7 117 54.9 112 85.3 114 58.0 113 53.6 113 69.8 108 54.6 113 52.8 115 57.9 113 61.8 118 46.5 113 77.6 113 48.9 113 31.3 37 54.3 46 38.9 61 37.8 120 50.6 120 49.7 120
2bit-BM-tele [98]106.0 82.4 119 87.2 114 91.8 123 44.4 101 54.7 90 52.9 102 46.0 104 68.3 100 47.4 99 51.5 111 54.4 91 53.4 112 40.5 96 47.1 84 47.7 105 47.9 116 66.4 96 53.1 114 71.7 121 83.6 123 75.1 122 21.9 99 39.3 106 29.2 95
UnFlow [129]107.6 89.6 124 94.1 122 86.4 116 72.2 121 83.9 124 77.9 120 71.4 121 93.2 123 69.8 119 54.8 114 74.4 111 51.4 110 62.0 122 66.9 123 69.1 125 50.0 118 80.9 114 57.1 117 53.2 100 69.0 96 7.68 4 15.6 79 30.8 80 21.0 80
Periodicity [78]108.9 68.9 107 83.5 101 65.5 98 52.2 108 69.8 112 57.0 105 78.4 123 82.5 112 87.2 125 47.2 106 74.7 112 45.5 104 69.7 127 81.9 127 65.6 122 59.5 122 84.9 124 60.7 121 36.3 50 79.8 117 19.2 21 40.7 122 66.5 126 53.1 122
Horn & Schunck [3]109.2 74.1 113 93.2 121 86.9 117 49.1 105 73.8 115 53.9 104 53.1 111 89.0 119 54.6 107 50.9 110 81.4 117 52.4 111 51.3 114 58.8 114 54.3 111 41.2 108 82.3 118 44.6 111 55.0 102 74.3 107 19.6 24 40.7 122 56.8 122 48.8 119
Heeger++ [104]111.4 86.1 121 91.0 118 77.1 104 67.6 120 91.6 126 65.1 110 85.6 126 94.6 125 83.6 124 71.0 118 88.2 120 68.8 116 62.9 123 69.8 125 67.3 123 69.4 126 91.0 126 70.8 124 40.2 66 80.9 121 26.3 44 21.7 96 31.2 83 27.4 89
SLK [47]112.4 67.5 105 90.3 117 82.1 111 72.2 121 84.7 125 84.8 122 58.4 116 94.0 124 58.1 114 78.1 121 82.5 118 84.6 123 55.4 120 61.5 119 68.1 124 49.4 117 83.7 122 57.7 118 36.2 49 68.1 94 26.2 43 50.3 125 60.0 123 65.2 126
FFV1MT [106]113.7 85.0 120 92.0 119 84.4 114 60.6 112 83.8 123 60.8 108 85.4 125 92.2 122 88.2 126 71.2 119 89.9 123 69.1 117 64.0 125 69.3 124 78.0 126 69.2 125 91.5 127 73.2 126 51.7 95 73.9 106 45.3 79 21.7 96 31.2 83 27.4 89
TI-DOFE [24]114.5 90.7 125 94.6 124 97.1 126 76.9 125 79.5 122 89.4 126 73.1 122 96.1 127 74.4 122 84.6 125 93.6 125 88.3 125 52.9 116 59.9 115 63.9 120 44.5 112 83.6 121 53.5 115 42.9 71 68.0 93 17.0 16 50.5 126 65.5 125 62.4 124
FOLKI [16]114.6 72.5 109 84.5 103 87.5 118 62.3 116 74.9 117 74.0 119 56.9 113 87.2 116 57.7 112 59.8 115 78.1 115 65.4 115 53.3 118 61.2 118 62.7 119 50.9 119 81.0 115 62.7 122 47.3 89 73.3 104 56.8 106 49.0 124 62.9 124 64.3 125
PGAM+LK [55]119.9 79.4 118 92.7 120 88.7 120 62.9 117 76.8 121 71.9 115 59.9 117 88.3 118 62.8 117 74.9 120 90.6 124 75.6 119 54.4 119 61.0 116 64.6 121 58.1 121 82.7 119 58.6 119 77.6 125 85.6 124 78.1 124 38.8 121 52.7 121 50.7 121
Adaptive flow [45]120.3 91.3 126 95.7 125 96.2 124 75.8 124 75.9 119 86.2 123 68.9 119 85.6 115 72.4 120 80.6 123 85.2 119 83.9 122 57.3 121 61.5 119 60.2 117 66.7 123 84.1 123 70.4 123 90.2 126 92.7 126 95.0 125 27.7 110 40.5 107 37.0 108
HCIC-L [99]121.1 88.6 123 96.7 127 89.3 121 80.8 126 74.4 116 93.1 127 86.6 127 87.2 116 95.2 127 95.8 127 97.4 127 96.9 127 63.4 124 66.7 122 57.8 116 68.9 124 82.7 119 73.1 125 96.0 127 95.9 127 98.6 126 25.3 108 33.9 93 34.7 105
Pyramid LK [2]123.3 86.7 122 94.5 123 96.2 124 73.1 123 76.5 120 86.4 125 70.8 120 89.6 120 78.5 123 78.1 121 88.6 122 82.7 121 68.8 126 76.4 126 80.4 127 75.7 127 85.1 125 78.1 127 73.1 122 79.6 116 69.3 118 60.8 127 74.5 127 79.9 127
AdaConv-v1 [126]128.0 100.0 128 100.0 128 100.0 128 100.0 128 100.0 128 100.0 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 100.0 128 99.7 128 99.9 128 99.9 128 99.9 128 99.9 128
SepConv-v1 [127]128.0 100.0 128 100.0 128 100.0 128 100.0 128 100.0 128 100.0 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 99.9 128 100.0 128 99.7 128 99.9 128 99.9 128 99.9 128 99.9 128
Move the mouse over the numbers in the table to see the corresponding images. Click to compare with the ground truth.

References

Methodtime*framescolor Reference and notes
[1] 2D-CLG 844 2 gray The 2D-CLG method by Bruhn et al. as implemented by Stefan Roth. [A. Bruhn, J. Weickert, and C. Schnörr. Lucas/Kanade meets Horn/Schunck: combining local and global optic flow methods. IJCV 63(3), 2005.] Parameters were set to match the published performance on Yosemite sequence, which may not be optimal for other sequences.
[2] Pyramid LK 12 2 color A modification of Bouguet's pyramidal implementation of Lucas-Kanade.
[3] Horn & Schunck 49 2 gray A modern Matlab implementation of the Horn & Schunck method by Deqing Sun. Parameters set to optimize AAE on all training data.
[4] Black & Anandan 328 2 gray A modern Matlab implementation of the Black & Anandan method by Deqing Sun.
[5] Brox et al. 18 2 color T. Brox, A. Bruhn, N. Papenberg, and J. Weickert. High accuracy optical flow estimation based on a theory for warping. ECCV 2004. (Improved using separate robust functions as proposed in A. Bruhn and J. Weickert, Towards ultimate motion estimation, ICCV 2005; improved by training on the training set.)
[6] Fusion 2,666 2 color V. Lempitsky, S. Roth, and C. Rother. Discrete-continuous optimization for optical flow estimation. CVPR 2008.
[7] Dynamic MRF 366 2 gray B. Glocker, N. Paragios, N. Komodakis, G. Tziritas, and N. Navab. Optical flow estimation with uncertainties through dynamic MRFs. CVPR 2008. (Method improved since publication.)
[8] Second-order prior 14 2 gray W. Trobin, T. Pock, D. Cremers, and H. Bischof. An unbiased second-order prior for high-accuracy motion estimation. DAGM 2008. (Method improved since publication; for details see W. Trobin, Ph.D. thesis, 2009.)
[9] GroupFlow 600 2 gray X. Ren. Local Grouping for Optical Flow. CVPR 2008.
[10] SegOF 60 2 color L. Xu, J. Chen, and J. Jia. Segmentation based variational model for accurate optical flow estimation. ECCV 2008. Code available.
[11] Learning Flow 825 2 gray D. Sun, S. Roth, J.P. Lewis, and M. Black. Learning optical flow (SRF-LFC). ECCV 2008.
[12] CBF 69 2 color W. Trobin, T. Pock, D. Cremers, and H. Bischof. Continuous energy minimization via repeated binary fusion. ECCV 2008. (Method improved since publication; for details see W. Trobin, Ph.D. thesis, 2009.)
[13] SPSA-learn 200 2 color Y. Li and D. Huttenlocher. Learning for optical flow using stochastic optimization. ECCV 2008.
[14] GraphCuts 1,200 2 color T. Cooke. Two applications of graph-cuts to image processing. DICTA 2008.
[15] F-TV-L1 8 2 gray A. Wedel, T. Pock, J. Braun, U. Franke, and D. Cremers. Duality TV-L1 flow with fundamental matrix prior. IVCNZ 2008.
[16] FOLKI 1.4 2 gray G. Le Besnerais and F. Champagnat. Dense optical flow by iterative local window registration. ICIP 2005.
[17] TV-L1-improved 2.9 2 gray A. Wedel, T. Pock, C. Zach, H. Bischof, and D. Cremers. An improved algorithm for TV-L1 optical flow computation. Proceedings of the Dagstuhl Visual Motion Analysis Workshop 2008. Code at GPU4Vision.
[18] DPOF 287 2 color C. Lei and Y.-H. Yang. Optical flow estimation on coarse-to-fine region-trees using discrete optimization. ICCV 2009. (Method improved since publication.)
[19] Filter Flow 34,000 2 color S. Seitz and S. Baker. Filter flow. ICCV 2009.
[20] Adaptive 9.2 2 gray A. Wedel, D. Cremers, T. Pock, and H. Bischof. Structure- and motion-adaptive regularization for high accuracy optic flow. ICCV 2009.
[21] Complementary OF 44 2 color H. Zimmer, A. Bruhn, J. Weickert, L. Valgaerts, A. Salgado, B. Rosenhahn, and H.-P. Seidel. Complementary optic flow. EMMCVPR 2009.
[22] Aniso. Huber-L1 2 2 gray M. Werlberger, W. Trobin, T. Pock, A. Wedel, D. Cremers, and H. Bischof. Anisotropic Huber-L1 optical flow. BMVC 2009. Code at GPU4Vision.
[23] Rannacher 0.12 2 gray J. Rannacher. Realtime 3D motion estimation on graphics hardware. Bachelor thesis, Heidelberg University, 2009.
[24] TI-DOFE 260 2 gray C. Cassisa, S. Simoens, and V. Prinet. Two-frame optical flow formulation in an unwarped multiresolution scheme. CIARP 2009.
[25] NL-TV-NCC 20 2 color M. Werlberger, T. Pock, and H. Bischof. Motion estimation with non-local total variation regularization. CVPR 2010.
[26] MDP-Flow 188 2 color L. Xu, J. Jia, and Y. Matsushita. Motion detail preserving optical flow estimation. CVPR 2010.
[27] ACK-Prior 5872 2 color K. Lee, D. Kwon, I. Yun, and S. Lee. Optical flow estimation with adaptive convolution kernel prior on discrete framework. CVPR 2010.
[28] LDOF 122 2 color T. Brox and J. Malik. Large displacement optical flow: descriptor matching in variational motion estimation. PAMI 33(3):500-513, 2011.
[29] p-harmonic 565 2 gray J. Gai and R. Stevenson. Optical flow estimation with p-harmonic regularization. ICIP 2010.
[30] TriangleFlow 4200 2 gray B. Glocker, H. Heibel, N. Navab, P. Kohli, and C. Rother. TriangleFlow: Optical flow with triangulation-based higher-order likelihoods. ECCV 2010.
[31] Classic+NL 972 2 color D. Sun, S. Roth, and M. Black. Secrets of optical flow estimation and their principles. CVPR 2010. Matlab code.
[32] Classic++ 486 2 gray A modern implementation of the classical formulation descended from Horn & Schunck and Black & Anandan; see D. Sun, S. Roth, and M. Black, Secrets of optical flow estimation and their principles, CVPR 2010.
[33] Nguyen 33 2 gray D. Nguyen. Tuning optical flow estimation with image-driven functions. ICRA 2011.
[34] Modified CLG 133 2 gray R. Fezzani, F. Champagnat, and G. Le Besnerais. Combined local global method for optic flow computation. EUSIPCO 2010.
[35] ComplOF-FED-GPU 0.97 2 color P. Gwosdek, H. Zimmer, S. Grewenig, A. Bruhn, and J. Weickert. A highly efficient GPU implementation for variational optic flow based on the Euler-Lagrange framework. CVGPU Workshop 2010.
[36] Ad-TV-NDC 35 2 gray M. Nawaz. Motion estimation with adaptive regularization and neighborhood dependent constraint. DICTA 2010.
[37] Layers++ 18206 2 color D. Sun, E. Sudderth, and M. Black. Layered image motion with explicit occlusions, temporal consistency, and depth ordering. NIPS 2010.
[38] OFH 620 3 color H. Zimmer, A. Bruhn, J. Weickert. Optic flow in harmony. IJCV 93(3) 2011.
[39] LSM 1615 2 color K. Jia, X. Wang, and X. Tang. Optical flow estimation using learned sparse model. ICCV 2011.
[40] CostFilter 55 2 color C. Rhemann, A. Hosni, M. Bleyer, C. Rother, and M. Gelautz. Fast cost-volume filtering for visual correspondence and beyond. CVPR 2011.
[41] Bartels 0.15 2 gray C. Bartels and G. de Haan. Smoothness constraints in recursive search motion estimation for picture rate conversion. IEEE TCSVT 2010. Version improved since publication: mapped on GPU.
[42] Shiralkar 600 2 gray M. Shiralkar and R. Schalkoff. A self organization-based optical flow estimator with GPU implementation. MVA 23(6):1229-1242.
[43] HBpMotionGpu 1000 5 gray S. Grauer-Gray and C. Kambhamettu. Hierarchical belief propagation to reduce search space using CUDA for stereo and motion estimation. WACV 2009. (Method improved since publication.)
[44] StereoFlow 7200 2 color G. Rosman, S. Shem-Tov, D. Bitton, T. Nir, G. Adiv, R. Kimmel, A. Feuer, and A. Bruckstein. Over-parameterized optical flow using a stereoscopic constraint. SSVM 2011:761-772.
[45] Adaptive flow 121 2 gray T. Arici. Energy minimization based motion estimation using adaptive smoothness priors. Submitted to IEEE TIP 2011.
[46] TC-Flow 2500 5 color S. Volz, A. Bruhn, L. Valgaerts, and H. Zimmer. Modeling temporal coherence for optical flow. ICCV 2011.
[47] SLK 300 2 gray T. Corpetti and E. Mémin. Stochastic uncertainty models for the luminance consistency assumption. IEEE TIP 2011.
[48] CLG-TV 29 2 gray M. Drulea. Total variation regularization of local-global optical flow. ITSC 2011. Matlab code.
[49] SimpleFlow 1.7 2 color M. Tao, J. Bai, P. Kohli, S. Paris. SimpleFlow: a non-iterative, sublinear optical flow algorithm. EUROGRAPHICS 2012.
[50] IAOF 57 2 gray D. Nguyen. Improving motion estimation using image-driven functions and hybrid scheme. PSIVT 2011.
[51] IAOF2 56 2 gray D. Nguyen. Enhancing the sharpness of flow field using image-driven functions with occlusion-aware filter. Submitted to TIP 2011.
[52] LocallyOriented 9541 2 gray Y.Niu, A. Dick, and M. Brooks. Locally oriented optical flow computation. To appear in TIP 2012.
[53] IROF-TV 261 2 color H. Rashwan, D. Puig, and M. Garcia. On improving the robustness of differential optical flow. ICCV 2011 Artemis workshop.
[54] Sparse Occlusion 2312 2 color A. Ayvaci, M. Raptis, and S. Soatto. Sparse occlusion detection with optical flow. Submitted to IJCV 2011.
[55] PGAM+LK 0.37 2 gray A. Alba, E. Arce-Santana, and M. Rivera. Optical flow estimation with prior models obtained from phase correlation. ISVC 2010.
[56] Sparse-NonSparse 713 2 color L. Chen, J. Wang, and Y. Wu. Decomposing and regularizing sparse/non-sparse components for motion field estimation. Submitted to PAMI 2013.
[57] nLayers 36150 4 color D. Sun, E. Sudderth, and M. Black. Layered segmentation and optical flow estimation over time. CVPR 2012.
[58] IROF++ 187 2 color H. Rashwan, D. Puig, and M. Garcia. Variational optical flow estimation based on stick tensor voting. IEEE TIP 2013.
[59] COFM 600 3 color M. Mozerov. Constrained optical flow estimation as a matching problem. IEEE TIP 2013.
[60] Efficient-NL 400 2 color P. Krähenbühl and V. Koltun. Efficient nonlocal regularization for optical flow. ECCV 2012.
[61] BlockOverlap 2 2 gray M. Santoro, G. AlRegib, and Y. Altunbasak. Motion estimation using block overlap minimization. Submitted to MMSP 2012.
[62] Ramp 1200 2 color A. Singh and N. Ahuja. Exploiting ramp structures for improving optical flow estimation. ICPR 2012.
[63] Occlusion-TV-L1 538 3 gray C. Ballester, L. Garrido, V. Lazcano, and V. Caselles. A TV-L1 optical flow method with occlusion detection. DAGM-OAGM 2012.
[64] TV-L1-MCT 90 2 color M. Mohamed and B. Mertsching. TV-L1 optical flow estimation with image details recovering based on modified census transform. ISVC 2012.
[65] Local-TV-L1 500 2 gray L. Raket. Local smoothness for global optical flow. ICIP 2012.
[66] ALD-Flow 61 2 color M. Stoll, A. Bruhn, and S. Volz. Adaptive integration of feature matches into variational optic flow methods. ACCV 2012.
[67] SIOF 234 2 color L. Xu, Z. Dai, and J. Jia. Scale invariant optical flow. ECCV 2012.
[68] MDP-Flow2 342 2 color L. Xu, J. Jia, and Y. Matsushita. Motion detail preserving optical flow estimation. PAMI 34(9):1744-1757, 2012. Code available.
[69] TCOF 1421 all gray J. Sanchez, A. Salgado, and N. Monzon. Optical flow estimation with consistent spatio-temporal coherence models. VISAPP 2013.
[70] LME 476 2 color W. Li, D. Cosker, M. Brown, and R. Tang. Optical flow estimation using Laplacian mesh energy. CVPR 2013.
[71] NN-field 362 2 color L. Chen, H. Jin, Z. Lin, S. Cohen, and Y. Wu. Large displacement optical flow from nearest neighbor fields. CVPR 2013.
[72] FESL 3310 2 color W. Dong, G. Shi, X. Hu, and Y. Ma. Nonlocal sparse and low-rank regularization for optical flow estimation. Submitted to IEEE TIP 2013.
[73] PMF 35 2 color J. Lu, H. Yang, D. Min, and M. Do. PatchMatch filter: efficient edge-aware filtering meets randomized search for fast correspondence field estimation. CVPR 2013.
[74] FC-2Layers-FF 2662 4 color D. Sun, J. Wulff, E. Sudderth, H. Pfister, and M. Black. A fully-connected layered model of foreground and background flow. CVPR 2013.
[75] Correlation Flow 290 2 color M. Drulea and S. Nedevschi. Motion estimation using the correlation transform. TIP 2013. Matlab code.
[76] TC/T-Flow 341 5 color M. Stoll, S. Volz, and A. Bruhn. Joint trilateral filtering for multiframe optical flow. ICIP 2013.
[77] OFLAF 1530 2 color T. Kim, H. Lee, and K. Lee. Optical flow via locally adaptive fusion of complementary data costs. ICCV 2013.
[78] Periodicity 8000 4 color G. Khachaturov, S. Gonzalez-Brambila, and J. Gonzalez-Trejo. Periodicity-based computation of optical flow. Submitted to Computacion y Sistemas (CyS) 2013.
[79] SILK 572 2 gray P. Zille, C. Xu, T. Corpetti, L. Shao. Observation models based on scale interactions for optical flow estimation. Submitted to IEEE TIP.
[80] CRTflow 13 3 color O. Demetz, D. Hafner, and J. Weickert. The complete rank transform: a tool for accurate and morphologically invariant matching of structures. BMVC 2013.
[81] SuperFlow 178 2 color Anonymous. Superpixel based optical flow estimation. ICCV 2013 submission 507.
[82] Aniso-Texture 300 2 color Anonymous. Texture information-based optical flow estimation using an incremental multi-resolution approach. ITC-CSCC 2013 submission 267.
[83] Classic+CPF 640 2 gray Z. Tu, R. Veltkamp, and N. van der Aa. A combined post-filtering method to improve accuracy of variational optical flow estimation. Submitted to Pattern Recognition 2013.
[84] S2D-Matching 1200 2 color Anonymous. Locally affine sparse-to-dense matching for motion and occlusion estimation. ICCV 2013 submission 1479.
[85] AGIF+OF 438 2 gray Z. Tu, R. Poppe, and R. Veltkamp. Adaptive guided image filter to warped interpolation image for variational optical flow computation. Submitted to Signal Processing 2015.
[86] DeepFlow 13 2 color P. Weinzaepfel, J. Revaud, Z. Harchaoui, and C. Schmid. DeepFlow: large displacement optical flow with deep matching. ICCV 2013.
[87] NNF-Local 673 2 color Z. Chen, H. Jin, Z. Lin, S. Cohen, and Y. Wu. Large displacement optical flow with nearest neighbor field. Submitted to PAMI 2014.
[88] EPPM w/o HM 2.5 2 color L. Bao, Q. Yang, and H. Jin. Fast edge-preserving PatchMatch for large displacement optical flow. CVPR 2014.
[89] MLDP_OF 165 2 gray M. Mohamed, H. Rashwan, B. Mertsching, M. Garcia, and D. Puig. Illumination-robust optical flow approach using local directional pattern. IEEE TCSVT 24(9):1499-1508, 2014.
[90] RFlow 20 2 gray S. Ali, C. Daul, and W. Blondel. Robust and accurate optical flow estimation for weak texture and varying illumination condition: Application to cystoscopy. IPTA 2014.
[91] SRR-TVOF-NL 32 all color P. Pohl, M. Sirotenko, E. Tolstaya, and V. Bucha. Edge preserving motion estimation with occlusions correction for assisted 2D to 3D conversion. IS&T/SPIE Electronic Imaging 2014.
[92] 2DHMM-SAS 157 2 color M.-C. Shih, R. Shenoy, and K. Rose. A two-dimensional hidden Markov model with spatially-adaptive states with application of optical flow. ICIP 2014 submission.
[93] WLIF-Flow 700 2 color Z. Tu, R. Veltkamp, N. van der Aa, and C. Van Gemeren. Weighted local intensity fusion method for variational optical flow estimation. Submitted to TIP 2014.
[94] FMOF 215 2 color N. Jith, A. Ramakanth, and V. Babu. Optical flow estimation using approximate nearest neighbor field fusion. ICASSP 2014.
[95] TriFlow 150 2 color TriFlow. Optical flow with geometric occlusion estimation and fusion of multiple frames. ECCV 2014 submission 914.
[96] ComponentFusion 6.5 2 color Anonymous. Fast optical flow by component fusion. ECCV 2014 submission 941.
[97] AggregFlow 1642 2 color D. Fortun, P. Bouthemy, and C. Kervrann. Aggregation of local parametric candidates and exemplar-based occlusion handling for optical flow. Preprint arXiv:1407.5759.
[98] 2bit-BM-tele 124 2 gray R. Xu and D. Taubman. Robust dense block-based motion estimation using a two-bit transform on a Laplacian pyramid. ICIP 2013.
[99] HCIC-L 330 2 color Anonymous. Globally-optimal image correspondence using a hierarchical graphical model. NIPS 2014 submission 114.
[100] TF+OM 600 2 color R. Kennedy and C. Taylor. Optical flow with geometric occlusion estimation and fusion of multiple frames. EMMCVPR 2015.
[101] PH-Flow 800 2 color J. Yang and H. Li. Dense, accurate optical flow estimation with piecewise parametric model. CVPR 2015.
[102] EpicFlow 16 2 color J. Revaud, P. Weinzaepfel, Z. Harchaoui, and C. Schmid. EpicFlow: edge-preserving interpolation of correspondences for optical flow. CVPR 2015.
[103] NNF-EAC 380 2 color Anonymous. Variational method for joint optical flow estimation and edge-aware image restoration. CVPR 2015 submission 2336.
[104] Heeger++ 6600 5 gray Anonymous. A context aware biologically inspired algorithm for optical flow (updated results). CVPR 2015 submission 2238.
[105] HBM-GC 330 2 color A. Zheng and Y. Yuan. Motion estimation via hierarchical block matching and graph cut. Submitted to ICIP 2015.
[106] FFV1MT 358 5 gray F. Solari, M. Chessa, N. Medathati, and P. Kornprobst. What can we expect from a V1-MT feedforward architecture for optical flow estimation? Submitted to Signal Processing: Image Communication 2015.
[107] ROF-ND 4 2 color S. Ali, C. Daul, E. Galbrun, and W. Blondel. Illumination invariant large displacement optical flow using robust neighbourhood descriptors. Submitted to CVIU 2015.
[108] DeepFlow2 16 2 color J. Revaud, P. Weinzaepfel, Z. Harchaoui, and C. Schmid. Deep convolutional matching. Submitted to IJCV, 2015.
[109] HAST 2667 2 color Anonymous. Highly accurate optical flow estimation on superpixel tree. ICCV 2015 submission 2221.
[110] FlowFields 15 2 color Anonymous. Flow Fields: Dense unregularized correspondence fields for highly accurate large displacement optical flow estimation. ICCV 2015 submission 744.
[111] SVFilterOh 1.56 2 color Anonymous. Fast estimation of large displacement optical flow using PatchMatch and dominant motion patterns. CVPR 2016 submission 1788.
[112] FlowNetS+ft+v 0.5 2 color Anonymous. Learning optical flow with convolutional neural networks. ICCV 2015 submission 235.
[113] CombBMOF 51 2 color M. Brüggemann, R. Kays, P. Springer, and O. Erdler. Combined block-matching and adaptive differential motion estimation in a hierarchical multi-scale framework. ICGIP 2014. (Method improved since publication.)
[114] PMMST 182 2 color F. Zhang, S. Xu, and X. Zhang. High accuracy correspondence field estimation via MST based patch matching. Submitted to TIP 2015.
[115] DF-Auto 70 2 color N. Monzon, A. Salgado, and J. Sanchez. Regularization strategies for discontinuity-preserving optical flow methods. Submitted to TIP 2015.
[116] CPM-Flow 3 2 color Anonymous. Efficient coarse-to-fine PatchMatch for large displacement optical flow. CVPR 2016 submission 241.
[117] CNN-flow-warp+ref 1.4 3 color D. Teney and M. Hebert. Learning to extract motion from videos in convolutional neural networks. ArXiv 1601.07532, 2016.
[118] Steered-L1 804 2 color Anonymous. Optical flow estimation via steered-L1 norm. Submitted to WSCG 2016.
[119] StereoOF-V1MT 343 2 gray Anonymous. Visual features for action-oriented tasks: a cortical-like model for disparity and optic flow computation. BMVC 2016 submission 132.
[120] PGM-C 5 2 color Y. Li. Pyramidal gradient matching for optical flow estimation. Submitted to PAMI 2016.
[121] RNLOD-Flow 1040 2 gray C. Zhang, Z. Chen, M. Wang, M. Li, and S. Jiang. Robust non-local TV-L1 optical flow estimation with occlusion detection. Submitted to TIP 2016.
[122] FlowNet2 0.091 2 color Anonymous. FlowNet 2.0: Evolution of optical flow estimation with deep networks. CVPR 2017 submission 900.
[123] S2F-IF 20 2 color Anonymous. S2F-IF: Slow-to-fast interpolator flow. CVPR 2017 submission 765.
[124] BriefMatch 0.068 2 gray G. Eilertsen, P.-E. Forssen, and J. Unger. Dense binary feature matching for real-time optical flow estimation. SCIA 2017 submission 62.
[125] OAR-Flow 60 2 color Anonymous. Order-adaptive regularisation for variational optical flow: global, local and in between. SSVM 2017 submission 20.
[126] AdaConv-v1 2.8 2 color S. Niklaus, L. Mai, and F. Liu. (Interpolation results only.) Video frame interpolation via adaptive convolution. CVPR 2017.
[127] SepConv-v1 0.2 2 color S. Niklaus, L. Mai, and F. Liu. (Interpolation results only.) Video frame interpolation via adaptive separable convolution. ICCV 2017.
[128] ProbFlowFields 37 2 color A. Wannenwetsch, M. Keuper, and S. Roth. ProbFlow: joint optical flow and uncertainty estimation. ICCV 2017.
[129] UnFlow 0.12 2 color Anonymous. UnFlow: Unsupervised learning of optical flow with a bidirectional census loss. Submitted to AAAI 2018.
* The "time" column lists the reported runtime in seconds on the "Urban" sequence. Note that these runtimes are not normalized by processor speed or type.