Andreas Geiger (MPI Tübingen) | Philip Lenz (KIT) | Christoph Stiller (KIT) | Raquel Urtasun (University of Toronto)

Optical Flow Evaluation 2012


The stereo / flow benchmark consists of 194 training image pairs and 195 test image pairs, saved in loss less png format. Our evaluation server computes the average number of bad pixels for all non-occluded or occluded (=all groundtruth) pixels. We require that all methods use the same parameter set for all test pairs. Our development kit provides details about the data format as well as MATLAB / C++ utility functions for reading and writing disparity maps and flow fields.

Our evaluation table ranks all methods according to the number of non-occluded erroneous pixels at the specified disparity / end-point error threshold. All methods providing less than 100 % density have been interpolated using simple background interpolation as explained in the corresponding header file in the development kit. For each method we show:

  • Out-Noc: Percentage of erroneous pixels in non-occluded areas
  • Out-All: Percentage of erroneous pixels in total
  • Avg-Noc: Average disparity / end-point error in non-occluded areas
  • Avg-All: Average disparity / end-point error in total
  • Density: Percentage of pixels for which ground truth has been provided by the method

Note: On 04.11.2013 we have improved the ground truth disparity maps and flow fields leading to slightly improvements for all methods. Please download the stereo/flow dataset with the improved ground truth for training again, if you have downloaded the dataset prior to 04.11.2013. Please consider reporting these new number for all future submissions. Links to last leaderboards before the updates: stereo and flow!

Important Policy Update: As more and more non-published work and re-implementations of existing work is submitted to KITTI, we have established a new policy: from now on, only submissions with significant novelty that are leading to a peer-reviewed paper in a conference or journal are allowed. Minor modifications of existing algorithms or student research projects are not allowed. Such work must be evaluated on a split of the training set. To ensure that our policy is adopted, new users must detail their status, describe their work and specify the targeted venue during registration. Furthermore, we will regularly delete all entries that are 6 months old but are still anonymous or do not have a paper associated with them. For conferences, 6 month is enough to determine if a paper has been accepted and to add the bibliography information. For longer review cycles, you need to resubmit your results.
Additional information used by the methods
  • Stereo: Method uses left and right (stereo) images
  • Multiview: Method uses more than 2 temporally adjacent images
  • Motion stereo: Method uses epipolar geometry for computing optical flow
  • Additional training data: Use of additional data sources for training (see details)

Error threshold        Evaluation area

Method Setting Code Out-Noc Out-All Avg-Noc Avg-All Density Runtime Environment
1 DM-Net-i2 code 0.00 % 0.00 % 0.0 px 0.0 px 0.00 % 0.90 s 1 core @ 2.5 Ghz (C/C++)
2 Anonym 0.00 % 0.00 % 0.0 px 0.0 px 0.00 % TBD s 1 core @ 2.5 Ghz (Python)
3 HD^3-Flow code 2.26 % 5.41 % 0.7 px 1.4 px 100.00 % 0.10 s NVIDIA Pascal Titan XP
Z. Yin, T. Darrell and F. Yu: Hierarchical Discrete Distribution Decomposition for Match Density Estimation. CVPR 2019.
4 PRSM
This method uses stereo information.
This method makes use of multiple (>2) views.
 code 2.46 % 4.23 % 0.7 px 1.0 px 100.00 % 300 s 1 core @ 2.5 Ghz (Matlab + C/C++)
C. Vogel, K. Schindler and S. Roth: 3D Scene Flow Estimation with a Piecewise Rigid Scene Model. ijcv 2015.
5 HTC 2.55 % 7.84 % 0.8 px 1.6 px 100.00 % 0.03 s 1 core @ 2.5 Ghz (C/C++)
6 LiteFlowNet2 code 2.72 % 6.30 % 0.7 px 1.4 px 100.00 % 0.0486 GTX 1080 (slower than Pascal Titan X)
T. Hui, X. Tang and C. Loy: A Lightweight Optical Flow CNN - Revisiting Data Fidelity and Regularization. arXiv preprint arXiv:1903.07414 2019.
7 VC-SF
This method uses stereo information.
This method makes use of multiple (>2) views.
 2.72 % 4.84 % 0.8 px 1.3 px 100.00 % 300 s 1 core @ 2.5 Ghz (Matlab + C/C++)
C. Vogel, S. Roth and K. Schindler: View-Consistent 3D Scene Flow Estimation over Multiple Frames. Proceedings of European Conference on Computer Vision. Lecture Notes in, Computer Science 2014.
8 SPS-StFl
This method uses stereo information.
This method makes use of the epipolar geometry.
 2.82 % 5.61 % 0.8 px 1.3 px 100.00 % 35 s 1 core @ 3.5 Ghz (C/C++)
K. Yamaguchi, D. McAllester and R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014.
9 LiteFlowNet2-MD+ 2.82 % 6.37 % 0.8 px 1.4 px 100.00 % 0.058 s NVIDIA GTX 1080
10 CompactFlow 2.98 % 6.54 % 0.8 px 1.5 px 100.00 % 0.05 s Nvidia Tesla V100 (Python)
11 SF-Net
This method uses stereo information.
 3.00 % 6.92 % 0.8 px 1.5 px 100.00 % 0.03 s GPU, GTX 1080Ti
12 3.07 % 6.89 % 0.8 px 1.5 px 100.00 %
13 LiteFlowNet code 3.27 % 7.27 % 0.8 px 1.6 px 100.00 % 0.0885 s GTX 1080 (slower than Pascal Titan X)
T. Hui, X. Tang and C. Loy: LiteFlowNet: A Lightweight Convolutional Neural Network for Optical Flow Estimation. Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2018.
14 SelFlow
This method makes use of multiple (>2) views.
 3.32 % 6.19 % 0.9 px 1.5 px 100.00 % 0.09 s NVIDIA GPU
P. Liu, M. Lyu, I. King and J. Xu: SelFlow: Self-Supervised Learning of Optical Flow. CVPR 2019.
15 PWC-Net+ code 3.36 % 6.72 % 0.8 px 1.4 px 100.00 % 0.03 s NVIDIA Pascal Titan X
D. Sun, X. Yang, M. Liu and J. Kautz: Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimation. arXiv preprint arXiv:1809.05571 2018.
16 SPS-Fl
This method makes use of the epipolar geometry.
 3.38 % 10.06 % 0.9 px 2.9 px 100.00 % 11 s 1 core @ 3.5 Ghz (C/C++)
K. Yamaguchi, D. McAllester and R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014.
17 OSF
This method uses stereo information.
 code 3.47 % 6.34 % 1.0 px 1.5 px 100.00 % 50 min 1 core @ 3.0 Ghz (Matlab + C/C++)
M. Menze and A. Geiger: Object Scene Flow for Autonomous Vehicles. Conference on Computer Vision and Pattern Recognition (CVPR) 2015.
18 PR-Sf+E
This method uses stereo information.
 3.57 % 7.07 % 0.9 px 1.6 px 100.00 % 200 s 4 cores @ 3.0 Ghz (Matlab + C/C++)
C. Vogel, K. Schindler and S. Roth: Piecewise Rigid Scene Flow. International Conference on Computer Vision (ICCV) 2013.
19 PCBP-Flow
This method makes use of the epipolar geometry.
 3.64 % 8.28 % 0.9 px 2.2 px 100.00 % 3 min 4 cores @ 2.5 Ghz (Matlab + C/C++)
K. Yamaguchi, D. McAllester and R. Urtasun: Robust Monocular Epipolar Flow Estimation. CVPR 2013.
20 SPN 3.71 % 7.87 % 0.9 px 1.7 px 100.00 % 0.08 s GPU @ 2.5 Ghz (C/C++)
21 FDFlowNet 3.72 % 8.21 % 0.9 px 1.7 px 100.00 % 0.02 s GPU @ 2.5 Ghz (Python)
22 PR-Sceneflow
This method uses stereo information.
 3.76 % 7.39 % 1.2 px 2.8 px 100.00 % 150 sec 4 core @ 3.0 Ghz (Matlab + C/C++)
C. Vogel, K. Schindler and S. Roth: Piecewise Rigid Scene Flow. International Conference on Computer Vision (ICCV) 2013.
23 SDF 3.80 % 7.69 % 1.0 px 2.3 px 100.00 % TBA s 1 core @ 2.5 Ghz (C/C++)
M. Bai*, W. Luo*, K. Kundu and R. Urtasun: Exploiting Semantic Information and Deep Matching for Optical Flow. ECCV 2016.
24 MotionSLIC
This method makes use of the epipolar geometry.
 3.91 % 10.56 % 0.9 px 2.7 px 100.00 % 11 s 1 core @ 3.0 Ghz (C/C++)
K. Yamaguchi, D. McAllester and R. Urtasun: Robust Monocular Epipolar Flow Estimation. CVPR 2013.
25 SfM-PM
This method makes use of multiple (>2) views.
 4.02 % 6.15 % 1.0 px 1.5 px 100.00 % 69 s 3 cores @ 3.6 Ghz (C/C++)
D. Maurer, N. Marniok, B. Goldluecke and A. Bruhn: Structure-from-Motion-Aware PatchMatch for Adaptive Optical Flow Estimation. ECCV 2018.
26 MFF
This method makes use of multiple (>2) views.
 4.19 % 7.87 % 0.9 px 1.7 px 100.00 % 0.05 s 1 core @ 2.5 Ghz (C/C++)
27 PWC-Net code 4.22 % 8.10 % 0.9 px 1.7 px 100.00 % 0.03 s NVIDIA Pascal Titan X
D. Sun, X. Yang, M. Liu and J. Kautz: PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volume. CVPR 2018.
28 ICCV-510 4.26 % 7.91 % 1.0 px 2.0 px 100.00 % 0.05 s 1 core @ 2.5 Ghz (Python)
29 UnFlow code 4.28 % 8.42 % 0.9 px 1.7 px 100.00 % 0.12 s GPU @ 1.5 Ghz (Python + C/C++)
S. Meister, J. Hur and S. Roth: UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss. AAAI 2018.
30 SelFlow
This method makes use of multiple (>2) views.
 4.31 % 7.68 % 1.0 px 2.2 px 100.00 % 0.09 s GPU @ 2.5 Ghz (Python)
P. Liu, M. Lyu, I. King and J. Xu: SelFlow: Self-Supervised Learning of Optical Flow. CVPR 2019.
31 MirrorFlow code 4.38 % 8.20 % 1.2 px 2.6 px 100.00 % 11 min 4 core @ 2.2 Ghz (C/C++)
J. Hur and S. Roth: MirrorFlow: Exploiting Symmetries in Joint Optical Flow and Occlusion Estimation. ICCV 2017.
32 ProFlow
This method makes use of multiple (>2) views.
 4.49 % 7.88 % 1.1 px 2.1 px 100.00 % 112 s GPU+CPU @ 3.6 Ghz (Python + C/C++)
D. Maurer and A. Bruhn: ProFlow: Learning to Predict Optical Flow. BMVC 2018.
33 HCF 4.56 % 9.17 % 1.0 px 1.9 px 100.00 % 0.02 s GPU @ 2.5 Ghz (Python)
34 DDFlow 4.57 % 8.86 % 1.1 px 3.0 px 100.00 % 0.06 s GPU @ >3.5 Ghz (Python)
P. Liu, I. King and M. Xu: DDFlow: Learning Optical Flow with Unlabeled Data Distillation. AAAI 2019.
35 ImpPB+SPCI code 4.65 % 13.47 % 1.1 px 2.9 px 100.00 % 60 s GPU @ 2.5 Ghz (Python)
T. Schuster, L. Wolf and D. Gadot: Optical Flow Requires Multiple Strategies (but only one network). CVPR 2017.
36 DIP-Flow-CPM
This method makes use of multiple (>2) views.
 4.69 % 9.63 % 1.0 px 2.4 px 100.00 % 52 s 2 cores @ 3.6 Ghz (C/C++)
D. Maurer, M. Stoll and A. Bruhn: Directional Priors for Multi-Frame Optical Flow. BMVC 2018.
37 CNNF+PMBP 4.70 % 14.87 % 1.1 px 3.3 px 100.00 % 30 min 1 core @ 3.5 Ghz (C/C++)
F. Zhang and B. Wah: Fundamental Principles on Learning New Features for Effective Dense Matching. IEEE Transactions on Image Processing 2018.
38 FlowNet2 code 4.82 % 8.80 % 1.0 px 1.8 px 100.00 % 0.1 s GPU @ 2.5 Ghz (C/C++)
E. Ilg, N. Mayer, T. Saikia, M. Keuper, A. Dosovitskiy and T. Brox: FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017.
39 FlowFieldCNN 4.89 % 13.01 % 1.2 px 3.0 px 100.00 % 23 s GPU @ 2.5 Ghz (C/C++)
C. Bailer, K. Varanasi and D. Stricker: CNN-based Patch Matching for Optical Flow with Thresholded Hinge Embedding Loss. CVPR 2017.
40 IntrpNt-df code 4.94 % 14.13 % 1.0 px 2.4 px 100.00 % 3 min GPU @ 2.5 Ghz (Python)
S. Zweig and L. Wolf: InterpoNet, a Brain Inspired Neural Network for Optical Flow Dense Interpolation. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017.
41 RicFlow 4.96 % 13.04 % 1.3 px 3.2 px 100.00 % 5 s 1 core @ 3.5 Ghz (C/C++)
Y. Hu, Y. Li and R. Song: Robust Interpolation of Correspondences for Large Displacement Optical Flow. CVPR 2017.
42 DIP-Flow-DF
This method makes use of multiple (>2) views.
 4.97 % 10.02 % 1.1 px 2.6 px 100.00 % 104s 2 cores @ 3.6 Ghz (C/C++)
D. Maurer, M. Stoll and A. Bruhn: Directional Priors for Multi-Frame Optical Flow. BMVC 2018.
43 EPIflow 4.97 % 11.21 % 1.1 px 3.2 px 100.00 % 0.1 s GPU @ 2.0 Ghz (Python)
44 FlowFields+ 5.06 % 13.14 % 1.2 px 3.0 px 100.00 % 28s 1 core @ 3.5 Ghz (C/C++)
C. Bailer, B. Taetz and D. Stricker: Flow Fields: Dense Correspondence Fields for Highly Accurate Large Displacement Optical Flow Estimation. .
45 deepmth 5.14 % 13.40 % 1.2 px 3.8 px 100.00 % 60 s 1 core @ 2.5 Ghz (C/C++)
46 DF+OIR 5.17 % 10.43 % 1.1 px 2.9 px 100.00 % 3 min 1 core @ 3.5 Ghz (Matlab + C/C++)
D. Maurer, M. Stoll and A. Bruhn: Order-Adaptive and Illumination Aware Variational Optical Flow Refinement. BMVC 2017.
47 UnOS(Full) 5.19 % 9.18 % 1.1 px 1.8 px 100.00 % 0.08 s GPU @ 2.5 Ghz (C/C++)
48 IntrpNt-cpm code 5.28 % 14.57 % 1.0 px 2.5 px 100.00 % 5.6 s GPU @ 2.5 Ghz (Python)
S. Zweig and L. Wolf: InterpoNet, a Brain Inspired Neural Network for Optical Flow Dense Interpolation. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017.
49 joint flow 5.28 % 10.44 % 1.1 px 2.3 px 100.00 % 0.08 s GPU @ 2.5 Ghz (C/C++)
50 PatchBatch code 5.29 % 14.17 % 1.3 px 3.3 px 100.00 % 50 s GPU @ 2.5 Ghz (Python)
D. Gadot and L. Wolf: PatchBatch: a Batch Augmented Loss for Optical Flow. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2016.
51 PCOF-SGBM
This method uses stereo information.
 5.40 % 8.73 % 1.2 px 2.1 px 100.00 % 0.8 s GPU @ 2.0 Ghz (C/C++)
M. Derome, A. Plyer, M. Sanfourche and G. Le Besnerais: A Prediction-Correction Approach for Real-Time Optical Flow Computation Using Stereo. German Conference on Pattern Recognition 2016.
52 SODA-Flow 5.57 % 10.71 % 1.3 px 2.8 px 100.00 % 96 s 2 cores @ 3.5 Ghz (C/C++)
D. Maurer, M. Stoll, S. Volz, P. Gairing and A. Bruhn: A Comparison of Isotropic and Anisotropic Second Order Regularisers for Optical Flow. SSVM 2017.
53 IntrpNt-ff code 5.57 % 14.76 % 1.1 px 2.6 px 100.00 % 25 s GPU @ 2.5 Ghz (Python)
S. Zweig and L. Wolf: InterpoNet, a Brain Inspired Neural Network for Optical Flow Dense Interpolation. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017.
54 MPIF 5.58 % 10.85 % 1.5 px 3.0 px 100.00 % 5 s 1 core @ 3.5 Ghz (C/C++)
55 PCOF
This method uses stereo information.
 5.59 % 9.69 % 1.2 px 1.9 px 100.00 % 0.08 s GPU @ 2.0 Ghz (C/C++)
M. Derome, A. Plyer, M. Sanfourche and G. Le Besnerais: A Prediction-Correction Approach for Real-Time Optical Flow Computation Using Stereo. German Conference on Pattern Recognition 2016.
56 CPM2 code 5.60 % 13.52 % 1.3 px 3.3 px 100.00 % 4 s 1 core @ 2.5 Ghz (C/C++)
57 OAR-Flow 5.69 % 10.72 % 1.4 px 2.8 px 100.00 % 90 s 2 cores @ 3.5 Ghz (C/C++)
D. Maurer, M. Stoll and A. Bruhn: Order-Adaptive Regularisation for Variational Optical Flow: Global, Local and in Between. SSVM 2017.
58 DDF code 5.73 % 14.18 % 1.4 px 3.4 px 100.00 % ~1 min GPU @ 2.5 Ghz (C/C++)
F. G\"uney and A. Geiger: Deep Discrete Flow. Asian Conference on Computer Vision (ACCV) 2016.
59 PH-Flow 5.76 % 10.57 % 1.3 px 2.9 px 100.00 % 800 s 1 core @ 3.5 Ghz (Matlab + C/C++)
J. Yang and H. Li: Dense, Accurate Optical Flow Estimation with Piecewise Parametric Model. CVPR 2015.
60 FlowFields code 5.77 % 14.01 % 1.4 px 3.5 px 100.00 % 23 s 4 cores @ 3.5 Ghz (C/C++)
C. Bailer, B. Taetz and D. Stricker: Flow Fields: Dense Correspondence Fields for Highly Accurate Large Displacement Optical Flow Estimation. International Conference on Computer Vision (ICCV) 2015.
61 CPM-Flow code 5.79 % 13.70 % 1.3 px 3.2 px 100.00 % 4.2s 1 core @ 3.5 Ghz (C/C++)
Y. Hu, R. Song and Y. Li: Efficient Coarse-to-Fine PatchMatch for Large Displacement Optical Flow. CVPR 2016.
62 IntrpNt-dm code 5.85 % 15.03 % 1.1 px 2.7 px 100.00 % 15 s GPU @ 2.5 Ghz (Python)
S. Zweig and L. Wolf: InterpoNet, a Brain Inspired Neural Network for Optical Flow Dense Interpolation. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017.
63 NLTGV-SC 5.93 % 11.96 % 1.6 px 3.8 px 100.00 % 16 s GPU @ 2.5 Ghz (Matlab + C/C++)
R. Ranftl, K. Bredies and T. Pock: Non-Local Total Generalized Variation for Optical Flow Estimation. Proceedings of the 13th European Conference on Computer Vision 2014.
64 TIMC 5.94 % 15.53 % 1.1 px 2.6 px 100.00 % 3.8 s GPU @ 2.5 Ghz (Matlab + C/C++)
65 DDS-DF 6.03 % 13.08 % 1.6 px 4.2 px 100.00 % 1 min 1 core @ 2.5 Ghz (Matlab + C/C++)
D. Wei, C. Liu and W. Freeman: A Data-driven Regularization Model for Stereo and Flow. 3DTV-Conference, 2014 International Conference on 2014.
66 TGV2ADCSIFT 6.20 % 15.15 % 1.5 px 4.5 px 100.00 % 12s GPU @ 2.4 Ghz (C/C++)
J. Braux-Zin, R. Dupont and A. Bartoli: A General Dense Image Matching Framework Combining Direct and Feature-based Costs. International Conference on Computer Vision (ICCV) 2013.
67 DiscreteFlow code 6.23 % 16.63 % 1.3 px 3.6 px 100.00 % 3 min 1 core @ 2.5 Ghz (Matlab + C/C++)
M. Menze, C. Heipke and A. Geiger: Discrete Optimization for Optical Flow. German Conference on Pattern Recognition (GCPR) 2015.
68 PPM code 6.23 % 15.91 % 1.4 px 5.0 px 100.00 % 36 s 1 core @ 2.8 Ghz (C/C++)
F. Kuang: PatchMatch algorithms for motion estimation and stereo reconstruction. 2017.
69 OD-Flow 6.42 % 12.65 % 1.4 px 2.9 px 100.00 % 0.3 s GPU @ 3.5 Ghz (Python)
70 UnOS(flownet-only) 6.46 % 14.33 % 1.2 px 3.2 px 100.00 % 0.08 s 1 core @ 2.5 Ghz (C/C++)
71 BTF-ILLUM 6.52 % 11.03 % 1.5 px 2.8 px 100.00 % 80 seconds 1 core @ 3.0 Ghz (C/C++)
O. Demetz, M. Stoll, S. Volz, J. Weickert and A. Bruhn: Learning Brightness Transfer Functions for the Joint Recovery of Illumination Changes and Optical Flow. Computer Vision -- ECCV 2014 2014.
72 DeepFlow2 code 6.61 % 17.35 % 1.4 px 5.3 px 100.00 % 22 s 1 core @ >3.5 Ghz (C/C++)
J. Revaud, P. Weinzaepfel, Z. Harchaoui and C. Schmid: DeepMatching: Hierarchical Deformable Dense Matching. 2015.
73 Devon 6.99 % 12.76 % 1.3 px 2.6 px 100.00 % 0.04 s 1 core @ 2.5 Ghz (C/C++)
74 Data-Flow code 7.11 % 14.57 % 1.9 px 5.5 px 100.00 % 3 min 2 cores @ 2.5 Ghz (Matlab + C/C++)
C. Vogel, S. Roth and K. Schindler: An Evaluation of Data Costs for Optical Flow. German Conference on Pattern Recognition (GCPR) 2013.
75 DeepFlow code 7.22 % 17.79 % 1.5 px 5.8 px 100.00 % 17 s 1 core @ 3.6Ghz (Python + C/C++)
P. Weinzaepfel, J. Revaud, Z. Harchaoui and C. Schmid: DeepFlow: Large displacement optical flow with deep matching. IEEE Intenational Conference on Computer Vision (ICCV) 2013.
76 EpicFlow code 7.88 % 17.08 % 1.5 px 3.8 px 100.00 % 15 s 1 core @ 3.6 Ghz (C/C++)
J. Revaud, P. Weinzaepfel, Z. Harchaoui and C. Schmid: EpicFlow: Edge-Preserving Interpolation of Correspondences for Optical Flow. CVPR 2015 - IEEE Conference on Computer Vision \& Pattern Recognition 2015.
77 TVL1-HOG 7.91 % 18.90 % 2.0 px 6.1 px 100.00 % 180 s 2 cores @ 3.0 Ghz (Matlab)
H. Rashwan, M. Mohamed, M. Garcia, B. Mertsching and D. Puig: Illumination Robust Optical Flow Model Based on Histogram of Oriented Gradients. German Conference on Pattern Recognition 2013 .
78 boostflow 8.38 % 12.54 % 1.3 px 2.4 px 100.00 % 0.08 s GPU @ 2.5 Ghz (C/C++)
79 MLDP-OF 8.67 % 18.78 % 2.4 px 6.7 px 100.00 % 160 s 2 cores @ 2.5 Ghz (Matlab)
M. Mohamed, H. Rashwan, B. Mertsching, M. Garcia and D. Puig: Illumination-Robust Optical Flow Using Local Directional Pattern . IEEE Transactions on Circuits and Systems for Video Technology 2014 .
80 SparseFlow code 9.09 % 19.32 % 2.6 px 7.6 px 100.00 % 10 s 1 core @ 3.5 Ghz (Matlab + C/C++)
R. Timofte and L. Gool: SparseFlow: Sparse Matching for Small to Large Displacement Optical Flow . WACV 2015 .
81 DeGraF-Flow 9.41 % 16.93 % 2.5 px 8.4 px 100.00 % 3.2 s 4 cores @ 2.5 Ghz (C/C++)
82 CRTflow 9.43 % 18.72 % 2.7 px 6.5 px 100.00 % 18 s GPU @ 1.0 Ghz (C/C++)
O. Demetz, D. Hafner and J. Weickert: The Complete Rank Transform: A Tool for Accurate and Morphologically Invariant Matching of Structure. Proc.~British Machine Vision Conference 2013 (BMVC) 2013.
83 C++ code 10.04 % 20.26 % 2.6 px 7.1 px 100.00 % 8.5 min 1 core @ 3.0 Ghz (Matlab)
D. Sun, S. Roth and M. Black: A Quantitative Analysis of Current Practices in Optical Flow Estimation and The Principles Behind Them. 2014.
84 TF+OFM
This method makes use of multiple (>2) views.
 code 10.22 % 18.46 % 2.0 px 5.0 px 100.00 % 350 s 1 cores @ 2.5 Ghz (Matlab + C/C++)
R. Kennedy and C. Taylor: Optical Flow with Geometric Occlusion Estimation and Fusion of Multiple Frames. EMMCVPR 2015.
85 boostflow_fnet 10.28 % 15.95 % 1.6 px 2.6 px 100.00 % 0.08 s GPU @ 2.5 Ghz (C/C++)
86 ROF-NND 10.44 % 21.23 % 2.5 px 6.5 px 100.00 % 50 s 4 cores @ 3.5 Ghz (Matlab + C/C++)
S. Ali, C. Daul, E. Galbrun and W. Blondel: Illumination invariant optical flow using neighborhood descriptors . Computer Vision and Image Understanding 2015.
87 C+NL code 10.49 % 20.64 % 2.8 px 7.2 px 100.00 % 14.8 min 1 core @ 3.0 Ghz (Matlab)
D. Sun, S. Roth and M. Black: A Quantitative Analysis of Current Practices in Optical Flow Estimation and The Principles Behind Them. 2014.
88 DSPyNet 10.64 % 19.10 % 2.4 px 5.5 px 100.00 % 0.02 s GPU @ 3.0 Ghz (C/C++)
Z. Sun and H. Wang: Deeper Spatial Pyramid Network with Refined Up-Sampling for Optical Flow Estimation. Proc. Pacific Rim Conference on Multimedia 2018.
89 fSGM 10.74 % 22.66 % 3.2 px 12.2 px 100.00 % 60 s 1 core @ 2.4 Ghz (C/C++)
S. Hermann and R. Klette: Hierarchical Scan Line Dynamic Programming for Optical Flow using Semi-Global Matching. ACCV Workshops 2012.
90 OD-Flow 10.76 % 18.92 % 1.7 px 4.4 px 100.00 % 0.3 s GPU @ 3.5 Ghz (Python)
91 TGV2CENSUS code 11.03 % 18.37 % 2.9 px 6.6 px 100.00 % 4 s GPU+CPU @ 3.0 Ghz (Matlab + C/C++)
M. Werlberger: Convex Approaches for High Performance Video Processing. 2012.
R. Ranftl, S. Gehrig, T. Pock and H. Bischof: Pushing the Limits of Stereo Using Variational Stereo Estimation. IV 2012.
92 PCA-Layers code 12.02 % 19.11 % 2.5 px 5.2 px 100.00 % 3.2 s 1 core @ 2.5 Ghz (Python + C/C++)
J. Wulff and M. Black: Efficient Sparse-to-Dense Optical Flow Estimation using a Learned Basis and Layers. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) 2015 2015.
93 AggregFlow 12.23 % 21.79 % 3.1 px 7.4 px 100.00 % 35 min 1 core @ 2.5 Ghz (C/C++)
D. Fortun, P. Bouthemy and C. Kervrann: Aggregation of local parametric candidates with exemplar-based occlusion handling for optical flow. Computer Vision and Image Understanding 2016.
94 SPyNet code 12.31 % 20.97 % 2.0 px 4.1 px 100.00 % 0.16 s Nvidia TitanX GPU (lua)
A. Ranjan and M. Black: Optical Flow Estimation using a Spatial Pyramid Network. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2017.
95 C+NL-fast code 12.36 % 22.28 % 3.2 px 7.9 px 100.00 % 2.9 min 1 core @ 3.0 Ghz (Matlab)
D. Sun, S. Roth and M. Black: A Quantitative Analysis of Current Practices in Optical Flow Estimation and The Principles Behind Them. 2014.
96 EPPM code 12.75 % 23.55 % 2.5 px 9.2 px 100.00 % 0.25 s GPU @ 1.0 Ghz (C/C++)
L. Bao, Q. Yang and H. Jin: Fast Edge-Preserving PatchMatch for Large Displacement Optical Flow. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2014.
97 CPNFlow 13.01 % 19.17 % 2.0 px 3.6 px 100.00 % 0.1 s GPU @ 1.5 Ghz (Python)
Y. Yang and S. Soatto: Conditional prior networks for optical flow. Proceedings of the European Conference on Computer Vision (ECCV) 2018.
98 HS code 14.75 % 24.11 % 4.0 px 9.0 px 100.00 % 2.6 min 1 core @ 3.0 Ghz (Matlab)
D. Sun, S. Roth and M. Black: A Quantitative Analysis of Current Practices in Optical Flow Estimation and The Principles Behind Them. 2014.
99 Grts-Flow-V2 15.63 % 26.41 % 3.2 px 8.4 px 100.00 % 0.3 s 1 core @ 1.5 Ghz (C/C++)
E. Zhu, Y. Li and Y. Shi: Fast Optical Flow Estimation Without Parallel Architectures. IEEE Transactions on Circuits and Systems for Video Technology 2016.
100 PCA-Flow code 15.67 % 24.59 % 2.7 px 6.2 px 100.00 % 0.19 s 1 core @ 2.5 Ghz (Python + C/C++)
J. Wulff and M. Black: Efficient Sparse-to-Dense Optical Flow Estimation using a Learned Basis and Layers. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) 2015 2015.
101 GC-BM-Bino
This method uses stereo information.
This method makes use of the epipolar geometry.
 18.83 % 29.30 % 5.0 px 12.1 px 83.73 % 1.3 s 2 cores @ 2.5 Ghz (C/C++)
B. Kitt and H. Lategahn: Trinocular Optical Flow Estimation for Intelligent Vehicle Applications. ITSC 2012.
102 eFolki 19.31 % 28.79 % 5.2 px 10.9 px 100.00 % 0.026 s GPU @ 700 Mhz (C/C++)
A. Plyer, G. Le Besnerais and F. Champagnat: Massively parallel Lucas Kanade optical flow for real-time video processing applications. Journal of Real-Time Image Processing 2014.
103 GC-BM-Mono
This method makes use of the epipolar geometry.
 19.38 % 29.80 % 5.0 px 12.1 px 84.33 % 1.3 s 2 cores @ 2.5 Ghz (C/C++)
B. Kitt and H. Lategahn: Trinocular Optical Flow Estimation for Intelligent Vehicle Applications. ITSC 2012.
104 UnOS(Ego) 20.23 % 25.25 % 2.3 px 3.1 px 100.00 % 0.08 s 1 core @ 2.5 Ghz (C/C++)
105 SVFilterOh 20.38 % 30.38 % 4.3 px 9.1 px 100.00 % 2 s 1 core @ 3 Ghz (C/C++), 1 GTX 780 GPU
M. Helala and F. Qureshi: Fast Estimation of Large Displacement Optical Flow Using Dominant Motion Patterns & Sub-Volume PatchMatch Filtering. Proc. 14th Conference on Computer and Robot Vision (CRV 17) 2017.
106 RSRS-Flow 20.78 % 29.75 % 6.2 px 12.1 px 100.00 % 4 min 1 core @ 2.5 Ghz (Matlab)
P. Ghosh and B. Manjunath: Robust Simultaneous Registration and Segmentation with Sparse Error Reconstruction. PAMI 2012.
107 Spynet-no-finetuning 20.88 % 30.28 % 4.7 px 10.0 px 100.00 % 0.5 s GPU @ >3.5 Ghz (Python)
108 ALD 21.37 % 30.71 % 10.9 px 16.0 px 100.00 % 110 s 1 core @ 2.5 Ghz (C/C++)
M. Stoll, S. Volz and A. Bruhn: Adaptive Integration of Feature Matches into Variational Optical Flow Methods. ACCV 2012.
109 LDOF code 21.93 % 31.39 % 5.6 px 12.4 px 100.00 % 1 min 1 core @ 2.5 Ghz (C/C++)
T. Brox and J. Malik: Large Displacement Optical Flow: Descriptor Matching in Variational Motion Estimation. PAMI 2011.
110 2Bit-BM-tele code 24.10 % 33.59 % 7.1 px 15.2 px 100.00 % 6 min 1 core @ 2.4 Ghz (C/C++)
R. Xu and D. Taubman: Robust Dense Block-Based Motion Estimation Using a Two-Bit Transform on a Laplacian Pyramid. 20th Proc. IEEE Int. Conf. Image Proc. 2013 2013.
111 DB-TV-L1 code 30.87 % 39.25 % 7.9 px 14.6 px 100.00 % 16 s 1 core @ 2.5 Ghz (Matlab)
C. Zach, T. Pock and H. Bischof: A Duality Based Approach for Realtime TV- L1 Optical Flow. DAGM 2007.
112 PyrLK code 31.48 % 40.08 % 15.6 px 29.6 px 92.33 % 1.3 s 4 cores @ 3.5 Ghz (C/C++)
J. Bouguet: Pyramidal implementation of the affine lucas kanade feature tracker description of the algorithm. .
113 GCSF
This method uses stereo information.
 33.17 % 41.71 % 7.0 px 15.3 px 48.27 % 2.4 s 1 core @ 2.5 Ghz (C/C++)
J. Cech, J. Sanchez-Riera and R. Horaud: Scene Flow Estimation by Growing Correspondence Seeds. CVPR 2011.
114 UnsupFlownet 34.85 % 43.15 % 4.6 px 11.3 px 100.00 % 0.03 s GPU @ 3.0 Ghz (C/C++)
J. Yu, A. Harley and K. Derpanis: Back to Basics: Unsupervised Learning of Optical Flow via Brightness Constancy and Motion Smoothness. ECCV 2016.
115 HAOF code 35.87 % 43.46 % 11.1 px 18.3 px 100.00 % 16.2 s 1 core @ 2.5 Ghz (C/C++)
T. Brox, A. Bruhn, N. Papenberg and J. Weickert: High accuracy optical flow estimation based on a theory for warping. ECCV 2004.
116 FSDEF
This method makes use of the epipolar geometry.
 36.85 % 44.65 % 8.8 px 16.4 px 41.81 % 0.26s 4 cores sandy bridge @ 3.5 Ghz (C/C++)
M. Garrigues and A. Manzanera: Fast Semi Dense Epipolar Flow Estimation. 2017 IEEE Winter Conference on Applications of Computer Vision (WACV) 2017.
117 FlowNetS+ft code 37.05 % 44.49 % 5.0 px 9.1 px 100.00 % 0.08 s GPU @ 1.0 Ghz (C/C++)
A. Dosovitskiy, P. Fischer, E. Ilg, P. Haeusser, C. Hazirbas, V. Golkov, P. Smagt, D. Cremers and T. Brox: FlowNet: Learning Optical Flow with Convolutional Networks. ICCV 2015.
118 RLOF(IM-GM) 37.49 % 44.78 % 8.2 px 15.4 px 11.84 % 3.7 s 4 core @ 3.4 Ghz (C/C++)
T. Senst, J. Geistert and T. Sikora: Robust local optical flow: Long-range motions and varying illuminations. 2016 IEEE International Conference on Image Processing (ICIP) 2016.
119 BERLOF code 37.66 % 45.27 % 8.5 px 16.2 px 15.26 % 0.231 s GPU @ 700 Mhz (C/C++) GeForce GTX 680
T. Senst, J. Geistert, I. Keller and T. Sikora: Robust Local Optical Flow Estimation using Bilinear Equations for Sparse Motion Estimation. 20th IEEE International Conference on Image Processing 2013.
120 DIS-FAST code 38.58 % 46.21 % 7.8 px 14.4 px 100.00 % 0.023 1 core @ 4 Ghz (C/C++)
T. Kroeger, R. Timofte, D. Dai and L. Van Gool: Fast Optical Flow using Dense Inverse Search. ECCV 2016.
121 RLOF code 38.60 % 46.13 % 8.7 px 16.5 px 14.76 % 0.488 s GPU @ 700 Mhz (C/C++) GeForce GTX 680
T. Senst, V. Eiselein and T. Sikora: Robust Local Optical Flow for Feature Tracking. TCSVT 2012.
122 Next-Flow code 39.12 % 46.37 % 5.1 px 9.2 px 100.00 % 0.1 s GPU @ 1.0 Ghz (C/C++)
N. Sedaghat, M. Zolfaghari and T. Brox: Hybrid Learning of Optical Flow and Next Frame Prediction to Boost Optical Flow in the Wild. 2017.
123 PolyExpand 47.59 % 54.00 % 17.3 px 25.3 px 100.00 % 1 s 1 core @ 2.5 Ghz (C/C++)
G. Farneback: Two-Frame Motion Estimation Based on Polynomial Expansion. SCIA 2003.
124 PMSFlow 57.20 % 62.41 % 10.3 px 17.5 px 100.00 % 0.08 s 1 core @ 2.5 Ghz (C/C++)
125 OCV-BM code 63.50 % 68.19 % 24.4 px 33.3 px 100.00 % 1.5 min 1 core @ 2.5 Ghz (C/C++)
G. Bradski: The OpenCV Library. Dr. Dobb's Journal of Software Tools 2000.
126 Pyramid-LK code 65.81 % 70.16 % 21.8 px 33.2 px 99.90 % 1.5 min 1 core @ 2.5 Ghz (Matlab)
J. Bouguet: Pyramidal implementation of the Lucas Kanade feature tracker. Intel 2000.
127 MEDIAN 79.37 % 82.46 % 16.0 px 24.0 px 99.94 % 0.01 s 1 core @ 2.5 Ghz (C/C++)
128 AVERAGE 81.27 % 84.06 % 16.3 px 24.7 px 99.94 % 0.01 s 1 core @ 2.5 Ghz (C/C++)
129 UFT 100.00 % 100.00 % 553.3 px 547.2 px 100.00 % 0.8 s GPU @ 1.0 Ghz (Python + C/C++)
This table as LaTeX


Related Datasets

  • HCI/Bosch Robust Vision Challenge: Optical flow and stereo vision challenge on high resolution imagery recorded at a high frame rate under diverse weather conditions (e.g., sunny, cloudy, rainy). The Robert Bosch AG provides a prize for the best performing method.
  • Image Sequence Analysis Test Site (EISATS): Synthetic image sequences with ground truth information provided by UoA and Daimler AG. Some of the images come with 3D range sensor information.
  • Middlebury Optical Flow Evaluation: The classic optical flow evaluation benchmark, featuring eight test images, with very accurate ground truth from a shape from UV light pattern system. 24 image pairs are provided in total.

Citation

When using this dataset in your research, we will be happy if you cite us:
@INPROCEEDINGS{Geiger2012CVPR,
  author = {Andreas Geiger and Philip Lenz and Raquel Urtasun},
  title = {Are we ready for Autonomous Driving? The KITTI Vision Benchmark Suite},
  booktitle = {Conference on Computer Vision and Pattern Recognition (CVPR)},
  year = {2012}
}



eXTReMe Tracker