The KITTI Vision Benchmark Suite (original) (raw)


The stereo 2015 / flow 2015 / scene flow 2015 benchmark consists of 200 training scenes and 200 test scenes (4 color images per scene, saved in loss less png format). Compared to the stereo 2012 and flow 2012 benchmarks, it comprises dynamic scenes for which the ground truth has been established in a semi-automatic process. Our evaluation server computes the percentage of bad pixels averaged over all ground truth pixels of all 200 test images. For this benchmark, we consider a pixel to be correctly estimated if the disparity or flow end-point error is <3px or <5% (for scene flow this criterion needs to be fulfilled for both disparity maps and the flow map). 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. More details can be found in Object Scene Flow for Autonomous Vehicles (CVPR 2015).

Our evaluation table ranks all methods according to the number of erroneous pixels. 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. Legend:

Note: On 13.03.2017 we have fixed several small errors in the flow (noc+occ) ground truth of the dynamic foreground objects and manually verified all images for correctness by warping them according to the ground truth. As a consequence, all error numbers have decreased slightly. Please download the devkit and the annotations with the improved ground truth for the training set again if you have downloaded the files prior to 13.03.2017 and consider reporting these new number in all future publications. The last leaderboards before these corrections can be found here (optical flow 2015) and here (scene flow 2015). The leaderboards for the KITTI 2015 stereo benchmarks did not change.

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.

Method Setting Code D1-bg D1-fg D1-all D2-bg D2-fg D2-all Fl-bg Fl-fg Fl-all SF-bg SF-fg SF-all Density Runtime Environment
1 MS-RAFT-3D 1.38 2.65 1.59 1.81 5.50 2.42 2.22 5.99 2.85 2.75 9.59 3.89 100.00 3 s GPU @ 2.5 Ghz (Python)
2 OAMaskFlow 1.48 3.46 1.81 1.91 7.52 2.84 2.07 7.11 2.91 2.66 10.90 4.03 100.00 0.5 s 1 core @ 2.5 Ghz (Python)
3 SplatFlow3D code 1.40 2.91 1.65 1.88 6.29 2.61 2.27 6.02 2.89 2.84 10.04 4.04 100.00 0.1 s 1 core @ 2.5 Ghz (Python)
B. Wang, Y. Zhang, J. Li, Y. Yu, Z. Sun, L. Liu and D. Hu: SplatFlow: Learning Multi-frame Optical Flow via Splatting. International Journal of Computer Vision 2024.
4 RFlow3D+gannet+cambg 1.48 3.46 1.81 1.90 7.94 2.91 2.08 6.95 2.89 2.63 11.91 4.17 100.00 0.1 s 1 core @ 2.5 Ghz (Python)
5 CamLiRAFT code 1.48 3.46 1.81 1.91 8.11 2.94 2.08 7.37 2.96 2.68 12.16 4.26 100.00 1 s GPU @ 2.5 Ghz (Python + C/C++)
H. Liu, T. Lu, Y. Xu, J. Liu and L. Wang: Learning Optical Flow and Scene Flow with Bidirectional Camera-LiDAR Fusion. TPAMI 2023.
6 ScaleFlow++RBO 1.48 3.46 1.81 1.92 8.29 2.98 2.31 6.39 2.99 2.91 11.34 4.32 100.00 0.1 s GPU @ 2.5 Ghz (Python)
7 CamLiFlow code 1.48 3.46 1.81 1.92 8.14 2.95 2.31 7.04 3.10 2.87 12.23 4.43 100.00 1.2 s GPU @ 2.5 Ghz (Python + C/C++)
H. Liu, T. Lu, Y. Xu, J. Liu, W. Li and L. Chen: CamLiFlow: Bidirectional Camera-LiDAR Fusion for Joint Optical Flow and Scene Flow Estimation. CVPR 2022.
8 RAFT-3D (CroCo) 1.38 2.65 1.59 2.15 7.20 2.99 2.91 7.41 3.66 3.43 10.94 4.68 100.00 1.5 s GPU @ 2.5 Ghz (Python)
9 M-FUSE This method makes use of multiple (>2) views. code 1.40 2.91 1.65 2.14 8.10 3.13 2.66 7.47 3.46 3.43 11.84 4.83 100.00 1.3 s GPU
L. Mehl, A. Jahedi, J. Schmalfuss and A. Bruhn: M-FUSE: Multi-frame Fusion for Scene Flow Estimation. Proc. Winter Conference on Applications of Computer Vision (WACV) 2023.
10 RigidMask+ISF code 1.53 3.65 1.89 2.09 8.92 3.23 2.63 7.85 3.50 3.25 13.08 4.89 100.00 3.3 s GPU @ 2.5 Ghz (Python)
G. Yang and D. Ramanan: Learning to Segment Rigid Motions from Two Frames. CVPR 2021.
11 PAFlow 1.48 3.46 1.81 2.04 7.94 3.02 2.75 6.86 3.43 3.61 11.79 4.97 100.00 0.53 s 1 core @ 2.5 Ghz (C/C++)
12 CamLiRAFT-NR code 1.48 3.46 1.81 2.05 7.86 3.02 2.76 6.78 3.43 3.64 11.66 4.97 100.00 1 s GPU @ 2.5 Ghz (Python + C/C++)
H. Liu, T. Lu, Y. Xu, J. Liu and L. Wang: Learning Optical Flow and Scene Flow with Bidirectional Camera-LiDAR Fusion. arXiv preprint arXiv:2303.12017 2023.
13 SF2SE3 code 1.40 2.91 1.65 2.20 7.66 3.11 3.17 8.79 4.11 3.75 13.15 5.32 100.00 2.7 s GPU @ >3.5 Ghz (Python)
L. Sommer, P. Schröppel and T. Brox: SF2SE3: Clustering Scene Flow into SE (3)-Motions via Proposal and Selection. DAGM German Conference on Pattern Recognition 2022.
14 RAFT-3D 1.48 3.46 1.81 2.51 9.46 3.67 3.39 8.79 4.29 4.27 13.27 5.77 100.00 2 s GPU @ 2.5 Ghz (Python + C/C++)
Z. Teed and J. Deng: RAFT-3D: Scene Flow using Rigid-Motion Embeddings. arXiv preprint arXiv:2012.00726 2020.
15 ScaleFlow++ 1.48 3.46 1.81 2.13 8.02 3.11 3.94 5.59 4.21 4.81 10.69 5.79 100.00 0.1 s GPU @ 2.5 Ghz (Python)
16 FP-TTC 1.48 3.46 1.81 2.43 9.14 3.54 4.23 5.71 4.48 5.13 11.35 6.16 100.00 0.15 s 1 core @ 2.5 Ghz (C/C++)
17 RFlow3D 2.51 6.04 3.09 3.22 9.41 4.25 3.04 6.77 3.66 4.90 13.39 6.31 100.00 0.1 s GPU @ 2.5 Ghz (Python)
18 UberATG-DRISF 2.16 4.49 2.55 2.90 9.73 4.04 3.59 10.40 4.73 4.39 15.94 6.31 100.00 0.75 s CPU+GPU @ 2.5 Ghz (Python)
W. Ma, S. Wang, R. Hu, Y. Xiong and R. Urtasun: Deep Rigid Instance Scene Flow. CVPR 2019.
19 Scale-flow code 1.48 3.46 1.81 2.55 8.24 3.50 5.24 5.71 5.32 6.06 11.32 6.94 100.00 0.8 s GPU @ 2.5 Ghz (Python)
H. Ling, Q. Sun, Z. Ren, Y. Liu, H. Wang and Z. Wang: Scale-flow: Estimating 3D Motion from Video. Proceedings of the 30th ACM International Conference on Multimedia 2022.
20 ACOSF 2.79 7.56 3.58 3.82 12.74 5.31 4.56 12.00 5.79 5.61 19.38 7.90 100.00 5 min 1 core @ 3.0 Ghz (Matlab + C/C++)
C. Li, H. Ma and Q. Liao: Two-Stage Adaptive Object Scene Flow Using Hybrid CNN-CRF Model. International Conference on Pattern Recognition (ICPR) 2020.
21 ISF 4.12 6.17 4.46 4.88 11.34 5.95 5.40 10.29 6.22 6.58 15.63 8.08 100.00 10 min 1 core @ 3 Ghz (C/C++)
A. Behl, O. Jafari, S. Mustikovela, H. Alhaija, C. Rother and A. Geiger: Bounding Boxes, Segmentations and Object Coordinates: How Important is Recognition for 3D Scene Flow Estimation in Autonomous Driving Scenarios?. International Conference on Computer Vision (ICCV) 2017.
22 Stereo expansion code 1.48 3.46 1.81 3.39 8.54 4.25 5.83 8.66 6.30 7.06 13.44 8.12 100.00 2 s GPU @ 2.5 Ghz (Python)
G. Yang and D. Ramanan: Upgrading Optical Flow to 3D Scene Flow through Optical Expansion. CVPR 2020.
23 ISDAFlow+ SAGFt 1.48 3.46 1.81 2.95 9.91 4.11 5.68 11.06 6.58 6.83 16.63 8.46 100.00 0.1 s 1 core @ 2.5 Ghz (C/C++)
24 Binary TTC 1.48 3.46 1.81 3.84 9.39 4.76 5.84 8.67 6.31 7.45 13.74 8.50 100.00 2 s GPU @ 1.0 Ghz (Python)
A. Badki, O. Gallo, J. Kautz and P. Sen: Binary TTC: A Temporal Geofence for Autonomous Navigation. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2021.
25 S3DGF 1.48 3.46 1.81 3.04 10.23 4.24 5.96 11.27 6.84 7.10 16.66 8.69 100.00 0.1 s GPU @ 2.5 Ghz (Python)
26 PRSM This method makes use of multiple (>2) views. code 3.02 10.52 4.27 5.13 15.11 6.79 5.33 13.40 6.68 6.61 20.79 8.97 99.99 300 s 1 core @ 2.5 Ghz (C/C++)
C. Vogel, K. Schindler and S. Roth: 3D Scene Flow Estimation with a Piecewise Rigid Scene Model. ijcv 2015.
27 DTF_SENSE This method makes use of multiple (>2) views. 2.08 3.13 2.25 4.82 9.02 5.52 7.31 9.48 7.67 8.21 14.08 9.18 100.00 0.76 s 1 core @ 2.5 Ghz (C/C++)
R. Schuster, C. Unger and D. Stricker: A Deep Temporal Fusion Framework for Scene Flow Using a Learnable Motion Model and Occlusions. IEEE Winter Conference on Applications of Computer Vision (WACV) 2021.
28 OSF+TC This method makes use of multiple (>2) views. 4.11 9.64 5.03 5.18 15.12 6.84 5.76 13.31 7.02 7.08 20.03 9.23 100.00 50 min 1 core @ 2.5 Ghz (C/C++)
M. Neoral and J. Šochman: Object Scene Flow with Temporal Consistency. 22nd Computer Vision Winter Workshop (CVWW) 2017.
29 SENSE code 2.07 3.01 2.22 4.90 10.83 5.89 7.30 9.33 7.64 8.36 15.49 9.55 100.00 0.32s GPU, GTX 2080Ti
H. Jiang, D. Sun, V. Jampani, Z. Lv, E. Learned-Miller and J. Kautz: SENSE: A Shared Encoder Network for Scene-Flow Estimation. The IEEE International Conference on Computer Vision (ICCV) 2019.
30 OSF 2018 code 4.11 11.12 5.28 5.01 17.28 7.06 5.38 17.61 7.41 6.68 24.59 9.66 100.00 390 s 1 core @ 2.5 Ghz (Matlab + C/C++)
M. Menze, C. Heipke and A. Geiger: Object Scene Flow. ISPRS Journal of Photogrammetry and Remote Sensing (JPRS) 2018.
31 SSF 3.55 8.75 4.42 4.94 17.48 7.02 5.63 14.71 7.14 7.18 24.58 10.07 100.00 5 min 1 core @ 2.5 Ghz (Matlab + C/C++)
Z. Ren, D. Sun, J. Kautz and E. Sudderth: Cascaded Scene Flow Prediction using Semantic Segmentation. International Conference on 3D Vision (3DV) 2017.
32 OSF code 4.54 12.03 5.79 5.45 19.41 7.77 5.62 18.92 7.83 7.01 26.34 10.23 100.00 50 min 1 core @ 2.5 Ghz (C/C++)
M. Menze and A. Geiger: Object Scene Flow for Autonomous Vehicles. Conference on Computer Vision and Pattern Recognition (CVPR) 2015.
33 ScaleFlow++_SAG 1.48 3.46 1.81 5.70 8.78 6.22 9.78 9.50 9.73 11.83 14.69 12.31 100.00 0.05 s GPU @ 2.5 Ghz (Python + C/C++)
34 DWARF 3.20 3.94 3.33 6.21 9.38 6.73 9.80 13.37 10.39 11.72 18.06 12.78 100.00 0.14s - 1.43s TitanXP - JetsonTX2
F. Aleotti, M. Poggi, F. Tosi and S. Mattoccia: Learning end-to-end scene flow by distilling single tasks knowledge. Thirty-Fourth AAAI Conference on Artificial Intelligence (AAAI-20) 2020.
35 ADFactory code 1.48 3.46 1.81 7.39 9.33 7.71 11.28 10.66 11.18 13.19 16.18 13.68 100.00 0.1 s GPU @ 2.5 Ghz (Python + C/C++)
H. Ling, Q. Sun, Y. Sun, X. Xu and X. Li: ADFactory: An Effective Framework for Generalizing Optical Flow with NeRF. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2024.
36 GS58_ScaleRES 1.48 3.46 1.81 8.11 9.02 8.26 11.09 10.77 11.04 14.02 16.18 14.38 100.00 0.1 s GPU @ 2.5 Ghz (Python + C/C++)
37 SFF++ This method makes use of multiple (>2) views. 4.27 12.38 5.62 7.31 18.12 9.11 10.63 17.48 11.77 12.44 25.33 14.59 100.00 78 s 4 cores @ 3.5 Ghz (C/C++)
R. Schuster, O. Wasenmüller, C. Unger, G. Kuschk and D. Stricker: SceneFlowFields++: Multi-frame Matching, Visibility Prediction, and Robust Interpolation for Scene Flow Estimation. International Journal of Computer Vision (IJCV) 2019.
38 DTF_PWOC This method makes use of multiple (>2) views. 3.91 8.57 4.68 6.25 14.03 7.55 10.78 19.99 12.31 12.42 25.74 14.64 100.00 0.38 s RTX 2080 Ti
R. Schuster, C. Unger and D. Stricker: A Deep Temporal Fusion Framework for Scene Flow Using a Learnable Motion Model and Occlusions. IEEE Winter Conference on Applications of Computer Vision (WACV) 2021.
39 Self-SuperFlow-ft 3.81 8.92 4.66 7.13 16.27 8.65 10.65 19.44 12.12 12.33 26.73 14.73 100.00 0.13 s GTX 1080 Ti
K. Bendig, R. Schuster and D. Stricker: Self-SuperFlow: Self-supervised Scene Flow Prediction in Stereo Sequences. International Conference on Image Processing (ICIP) 2022.
40 FSF+MS This method makes use of the epipolar geometry.This method makes use of multiple (>2) views. 5.72 11.84 6.74 7.57 21.28 9.85 8.48 25.43 11.30 11.17 33.91 14.96 100.00 2.7 s 4 cores @ 3.5 Ghz (C/C++)
T. Taniai, S. Sinha and Y. Sato: Fast Multi-frame Stereo Scene Flow with Motion Segmentation. IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017) 2017.
41 PWOC-3D code 4.19 9.82 5.13 7.21 14.73 8.46 12.40 15.78 12.96 14.30 22.66 15.69 100.00 0.13 s GTX 1080 Ti
R. Saxena, R. Schuster, O. Wasenmüller and D. Stricker: PWOC-3D: Deep Occlusion-Aware End-to-End Scene Flow Estimation. Intelligent Vehicles Symposium (IV) 2019.
42 CSF 4.57 13.04 5.98 7.92 20.76 10.06 10.40 25.78 12.96 12.21 33.21 15.71 99.99 80 s 1 core @ 2.5 Ghz (C/C++)
Z. Lv, C. Beall, P. Alcantarilla, F. Li, Z. Kira and F. Dellaert: A Continuous Optimization Approach for Efficient and Accurate Scene Flow. European Conf. on Computer Vision (ECCV) 2016.
43 SceneFFields 5.12 13.83 6.57 8.47 21.83 10.69 10.58 24.41 12.88 12.48 32.28 15.78 100.00 65 s 4 cores @ 3.7 Ghz (C/C++)
R. Schuster, O. Wasenmüller, G. Kuschk, C. Bailer and D. Stricker: SceneFlowFields: Dense Interpolation of Sparse Scene Flow Correspondences. IEEE Winter Conference on Applications of Computer Vision (WACV) 2018.
44 PR-Sceneflow code 4.74 13.74 6.24 11.14 20.47 12.69 11.73 24.33 13.83 13.49 31.22 16.44 100.00 150 s 4 core @ 3.0 Ghz (Matlab + C/C++)
C. Vogel, K. Schindler and S. Roth: Piecewise Rigid Scene Flow. ICCV 2013.
45 RFlow3D+monster 2.51 6.04 3.09 9.31 41.00 14.59 3.04 6.77 3.66 11.15 44.81 16.75 100.00 0.1 s 1 core @ 2.5 Ghz (C/C++)
46 GS58_Scale 1.48 3.46 1.81 13.67 9.37 12.95 11.22 11.76 11.31 18.46 17.25 18.26 100.00 0.1 s 1 core @ 2.5 Ghz (Python)
47 EMR-MSF 8.61 15.15 9.70 11.73 28.43 14.51 9.86 22.27 11.93 15.93 38.78 19.74 100.00 0.25 s GPU @ 2.5 Ghz (Python)
Z. Jiang and M. Okutomi: EMR-MSF: Self-Supervised Recurrent Monocular Scene Flow Exploiting Ego-Motion Rigidity. Proceedings of the IEEE/CVF International Conference on Computer Vision 2023.
48 GS_ScaleFlow 1.48 3.46 1.81 11.39 13.31 11.71 15.29 15.64 15.35 19.72 22.77 20.23 100.00 0.1 s GPU @ 2.5 Ghz (Python)
49 SPS+FF++ code 5.47 12.19 6.59 13.06 20.83 14.35 15.91 20.27 16.64 18.98 29.51 20.73 100.00 36 s 1 core @ 3.5 Ghz (C/C++)
R. Schuster, O. Wasenmüller and D. Stricker: Dense Scene Flow from Stereo Disparity and Optical Flow. ACM Computer Science in Cars Symposium (CSCS) 2018.
50 Mono-SF 14.21 26.94 16.32 16.89 33.07 19.59 11.40 19.64 12.77 19.79 39.57 23.08 100.00 41 s 1 core @ 3.5 Ghz (Matlab + C/C++)
F. Brickwedde, S. Abraham and R. Mester: Mono-SF: Multi-View Geometry meets Single-View Depth for Monocular Scene Flow Estimation of Dynamic Traffic Scenes. Proc. of International Conference on Computer Vision (ICCV) 2019.
51 SGM+SF 5.15 15.29 6.84 14.10 23.13 15.60 20.91 25.50 21.67 23.09 34.46 24.98 100.00 45 min 16 core @ 3.2 Ghz (C/C++)
H. Hirschmüller: Stereo Processing by Semiglobal Matching and Mutual Information. PAMI 2008.M. Hornacek, A. Fitzgibbon and C. Rother: SphereFlow: 6 DoF Scene Flow from RGB-D Pairs. CVPR 2014.
52 MonoComb 17.89 21.16 18.44 22.34 25.85 22.93 5.84 8.67 6.31 27.06 33.55 28.14 100.00 0.58 s RTX 2080 Ti
R. Schuster, C. Unger and D. Stricker: MonoComb: A Sparse-to-Dense Combination Approach for Monocular Scene Flow. ACM Computer Science in Cars Symposium (CSCS) 2020.
53 Self-SuperFlow 5.78 19.76 8.11 19.88 30.03 21.57 22.70 28.55 23.67 26.31 40.72 28.71 100.00 0.13 s GTX 1080 Ti
K. Bendig, R. Schuster and D. Stricker: Self-SuperFlow: Self-supervised Scene Flow Prediction in Stereo Sequences. International Conference on Image Processing (ICIP) 2022.
54 PCOF-LDOF 6.31 19.24 8.46 19.09 30.54 20.99 14.34 38.32 18.33 25.26 49.39 29.27 100.00 50 s 1 core @ 3.0 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.
55 PCOF + ACTF 6.31 19.24 8.46 19.15 36.27 22.00 14.89 60.15 22.43 25.77 67.75 32.76 100.00 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 Multi-Mono-SF-ft This method makes use of multiple (>2) views. code 21.60 28.22 22.71 25.47 31.72 26.51 12.41 18.20 13.37 31.18 42.68 33.09 100.00 0.06 s NVIDIA GTX 1080 Ti
J. Hur and S. Roth: Self-Supervised Multi-Frame Monocular Scene Flow. CVPR 2021.
57 SGM&FlowFie+ 11.93 20.57 13.37 27.02 31.71 27.80 22.83 22.75 22.82 32.26 40.12 33.57 81.24 29 s 1 core @ 3.5 Ghz (C/C++)
R. Schuster, C. Bailer, O. Wasenmüller and D. Stricker: Combining Stereo Disparity and Optical Flow for Basic Scene Flow. Commercial Vehicle Technology Symposium (CVTS) 2018.
58 Self-Mono-SF-ft code 20.72 29.41 22.16 23.83 32.29 25.24 15.51 17.96 15.91 31.51 45.77 33.88 100.00 0.09 s NVIDIA GTX 1080 Ti
J. Hur and S. Roth: Self-Supervised Monocular Scene Flow Estimation. CVPR 2020.
59 SGM+C+NL code 5.15 15.29 6.84 28.77 25.65 28.25 34.24 42.46 35.61 38.21 50.95 40.33 100.00 4.5 min 1 core @ 2.5 Ghz (C/C++)
H. Hirschmüller: Stereo Processing by Semiglobal Matching and Mutual Information. PAMI 2008.D. Sun, S. Roth and M. Black: A Quantitative Analysis of Current Practices in Optical Flow Estimation and the Principles Behind Them. IJCV 2013.
60 3DG-DVO 14.12 18.68 14.88 27.93 32.27 28.65 31.10 55.44 35.15 38.28 63.69 42.51 100.00 0.04 s GPU @ 1.5 Ghz (Python)
61 SGM+LDOF code 5.15 15.29 6.84 29.58 23.48 28.56 40.81 31.92 39.33 43.99 42.09 43.67 100.00 86 s 1 core @ 2.5 Ghz (C/C++)
H. Hirschmüller: Stereo Processing by Semiglobal Matching and Mutual Information. PAMI 2008.T. Brox and J. Malik: Large Displacement Optical Flow: Descriptor Matching in Variational Motion Estimation. PAMI 2011.
62 Multi-Mono-SF This method makes use of multiple (>2) views. code 27.48 47.30 30.78 32.39 44.56 34.41 18.13 26.59 19.54 40.29 62.78 44.04 100.00 0.06 s NVIDIA GTX 1080 Ti
J. Hur and S. Roth: Self-Supervised Multi-Frame Monocular Scene Flow. CVPR 2021.
63 DWBSF 19.61 22.69 20.12 35.72 28.15 34.46 40.74 31.16 39.14 46.42 40.76 45.48 100.00 7 min 4 cores @ 3.5 Ghz (C/C++)
C. Richardt, H. Kim, L. Valgaerts and C. Theobalt: Dense Wide-Baseline Scene Flow From Two Handheld Video Cameras. 3DV 2016.
64 Self-Mono-SF code 31.22 48.04 34.02 34.89 43.59 36.34 23.26 24.93 23.54 46.68 63.82 49.54 100.00 0.09 s NVIDIA GTX 1080 Ti
J. Hur and S. Roth: Self-Supervised Monocular Scene Flow Estimation. CVPR 2020.
65 GCSF code 11.64 27.11 14.21 32.94 35.77 33.41 47.38 41.50 46.40 52.92 56.68 53.54 100.00 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.
66 VSF code 27.31 21.72 26.38 59.51 44.93 57.08 50.06 45.40 49.28 67.69 62.93 66.90 100.00 125 min 1 core @ 2.5 Ghz (C/C++)
F. Huguet and F. Devernay: A Variational Method for Scene Flow Estimation from Stereo Sequences. ICCV 2007.
67 Stereo-RSSF code 56.60 73.05 59.34 58.86 74.41 61.45 70.68 73.60 71.17 76.21 81.62 77.11 9.26 2.5 s 8 core @ 2.5 Ghz (Matlab)
E. Salehi, A. Aghagolzadeh and R. Hosseini: Stereo-RSSF: stereo robust sparse scene-flow estimation. The Visual Computer 2023.