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

Stereo Evaluation 2015


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:

  • D1: Percentage of stereo disparity outliers in first frame
  • D2: Percentage of stereo disparity outliers in second frame
  • Fl: Percentage of optical flow outliers
  • SF: Percentage of scene flow outliers (=outliers in either D0, D1 or Fl)
  • bg: Percentage of outliers averaged only over background regions
  • fg: Percentage of outliers averaged only over foreground regions
  • all: Percentage of outliers averaged over all ground truth pixels


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.
Additional information used by the methods
  • Flow: Method uses optical flow (2 temporally adjacent 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)

Evaluation ground truth        Evaluation area

Method Setting Code D1-bg D1-fg D1-all Density Runtime Environment
1 M2S_CSPN 1.51 % 2.88 % 1.74 % 100.00 % 0.5 s GPU @ 2.5 Ghz (C/C++)
X. Cheng, P. Wang and R. Yang: Learning Depth with Convolutional Spatial Propagation Network. arXiv preprint arXiv:1810.02695 2018.
2 FBA-AMNet 1.53 % 3.43 % 1.84 % 100.00 % 0.9 s GPU @ 2.5 Ghz (Python)
3 CMSM 1.51 % 3.80 % 1.89 % 100.00 % 0.48 s GPU @ 2.5 Ghz (Python)
4 Samsung_System_LSI 1.56 % 3.56 % 1.90 % 100.00 % 0.4 s GPU @ 2.5 Ghz (Python)
5 RawStereoNet 1.57 % 3.56 % 1.90 % 100.00 % 0.43 s NVIDIA TITAN X Pascal (PyTorch)
6 MS_CSPN 1.56 % 3.78 % 1.93 % 100.00 % 0.5 s GPU @ 2.5 Ghz (C/C++)
X. Cheng, P. Wang and R. Yang: Learning Depth with Convolutional Spatial Propagation Network. arXiv preprint arXiv:1810.02695 2018.
7 GANet-15 1.55 % 3.82 % 1.93 % 100.00 % 0.36 s GPU (Pytorch)
8 NCA-Net 1.68 % 3.28 % 1.94 % 100.00 % 0.5 s GPU @ 2.5 Ghz (Python)
9 PSMNet_R 1.62 % 3.79 % 1.98 % 100.00 % 0.5 s GPU @ 2.5 Ghz (Python)
10 EdgeStereo-V2 1.84 % 3.30 % 2.08 % 100.00 % 0.32s Nvidia GTX Titan Xp
X. Song, X. Zhao, H. Hu and L. Fang: EdgeStereo: A Context Integrated Residual Pyramid Network for Stereo Matching. Asian Conference on Computer Vision 2018.
11 DSHNet 1.65 % 4.29 % 2.09 % 100.00 % 0.7 s Nvidia GTX Titan Xp
12 EMCUA 1.66 % 4.27 % 2.09 % 100.00 % 0.9 s 1 core @ 2.5 Ghz (C/C++)
13 KesonStereo_V1 1.77 % 3.74 % 2.09 % 100.00 % 0.4 s GPU @ 2.5 Ghz (Python)
14 open-depth 1.76 % 3.84 % 2.10 % 100.00 % 0.51 s NVIDIA TITAN Xp (PyTorch 0.4.0)
15 GwcNet-g 1.74 % 3.93 % 2.11 % 100.00 % 0.32 s GPU @ 2.0 Ghz (Python + C/C++)
16 SSPCV-Net 1.75 % 3.89 % 2.11 % 100.00 % 0.9 s GPU @ 2.5 Ghz (Python)
17 MS-Net 1.72 % 4.08 % 2.11 % 100.00 % 0.75 s 1 core @ 2.5 Ghz (C/C++)
18 PANet 1.79 % 3.75 % 2.12 % 100.00 % 0.6 s GPU @ 2.5 Ghz (Python)
19 IPSMNet 1.72 % 4.11 % 2.12 % 100.00 % 0.5 s GPU @ 2.5 Ghz (python)
20 DM-Net 1.69 % 4.29 % 2.12 % 100.00 % 0.9s 1 core @ 2.5 Ghz (Python)
21 MCUA 1.69 % 4.38 % 2.14 % 100.00 % 0.40s Titan XP
22 HSM-1.8x 1.80 % 3.85 % 2.14 % 100.00 % 0.14 s GPU @ 2.5 Ghz (Python)
23 Stereo-fusion-SJTU 1.87 % 3.61 % 2.16 % 100.00 % 0.7 s Nvidia GTX Titan Xp
X. Song, X. Zhao, H. Hu and L. Fang: EdgeStereo: A Context Integrated Residual Pyramid Network for Stereo Matching. Asian Conference on Computer Vision 2018.
24 SLED-Net 1.85 % 4.15 % 2.23 % 100.00 % 0.75 s 1 core @ 2.5 Ghz (C/C++)
25 TinyStereo_V2 1.93 % 3.76 % 2.24 % 100.00 % 0.4 s GPU @ 2.5 Ghz (Python)
26 SegStereo code 1.88 % 4.07 % 2.25 % 100.00 % 0.6 s Nvidia GTX Titan Xp
G. Yang, H. Zhao, J. Shi, Z. Deng and J. Jia: SegStereo: Exploiting Semantic Information for Disparity Estimation. ECCV 2018.
27 NLCA-Net 1.91 % 3.94 % 2.25 % 100.00 % 0.6 s 1 core @ 2.5 Ghz (C/C++)
28 HDU-LJJ-Group 1.82 % 4.42 % 2.25 % 100.00 % 0.47 s GPU @ 1.5 Ghz (Python)
29 Stereo-DRNet 1.72 % 4.95 % 2.26 % 100.00 % 0.23 s GPU@2.5hz
30 PASM 1.78 % 4.64 % 2.26 % 100.00 % 0.52 s 1 core @ 2.5 Ghz (C/C++)
31 MSDC-Net 1.96 % 3.77 % 2.26 % 100.00 % 0.6 s GPU @ 2.5 Ghz (Python)
32 MCV-MFC 1.95 % 3.84 % 2.27 % 100.00 % 0.35 s 1 core @ 2.5 Ghz (C/C++)
33 HSM-1.5x 1.95 % 3.93 % 2.28 % 100.00 % 0.085 s GPU @ 2.5 Ghz (Python)
34 TinyStereo 1.92 % 4.13 % 2.28 % 100.00 % 0.39 s 1 core @ 2.5 Ghz (C/C++)
35 CFP-Net 1.90 % 4.39 % 2.31 % 100.00 % 0.9 s 8 cores @ 2.5 Ghz (Python)
36 PSMNet code 1.86 % 4.62 % 2.32 % 100.00 % 0.41 s Nvidia GTX Titan Xp
J. Chang and Y. Chen: Pyramid Stereo Matching Network. arXiv preprint arXiv:1803.08669 2018.
37 SF-Net
This method uses optical flow information.
 2.17 % 3.13 % 2.33 % 100.00 % 0.21 s GPU, GTX 1080Ti
38 CAR 1.92 % 4.43 % 2.34 % 100.00 % 0.39 s GPU @ 2.5 Ghz (Python)
39 SE-PSM 1.90 % 4.59 % 2.34 % 100.00 % 0.85 s GPU @ 3.0 Ghz (Python)
40 disparity stereo 1.85 % 4.86 % 2.35 % 100.00 % 0.5 s GPU @ 1.5 Ghz (Python)
41 SWNet 1.92 % 4.60 % 2.36 % 100.00 % 0.4 s GPU @ 2.5 Ghz (Python)
42 RawStereoNet-r 1.87 % 4.86 % 2.37 % 100.00 % 0.43 s NVIDIA TITAN X Pascal (PyTorch）
43 DeepStereo_V2 2.00 % 4.21 % 2.37 % 100.00 % 0.4 s 1 core @ 2.5 Ghz (C/C++)
44 UMC-GAN 1.95 % 4.57 % 2.38 % 100.00 % 0.7 s GPU @ 2.5 Ghz (Python)
45 L2-method 1.91 % 4.90 % 2.40 % 100.00 % 0.35 s GPU @ 2.5 Ghz (Python + C/C++)
46 PSM+NN 1.95 % 4.85 % 2.43 % 100.00 % 1 s GPU @ 2.5 Ghz (Python + C/C++)
47 LWSM 1.86 % 5.35 % 2.44 % 100.00 % 0.24 s GPU @ 2.5 Ghz (Python)
48 RAP 2.00 % 4.83 % 2.47 % 100.00 % 0.54 s 1 core @ 2.5 Ghz (C/C++)
49 NNet 1.95 % 5.32 % 2.51 % 100.00 % 0.69 s GPU @ 2.5 Ghz (Python + C/C++)
50 CAR 2.05 % 4.81 % 2.51 % 100.00 % 0.3 s 1 core @ 2.5 Ghz (C/C++)
51 X_ASPP 2.13 % 4.57 % 2.54 % 100.00 % 0.88 s GPU @ 2.5 Ghz (Python)
52 DSSF
This method uses optical flow information.
 2.16 % 4.49 % 2.55 % 100.00 % 0.75 s CPU+GPU @ 2.5 Ghz (Python)
53 FBW-Net 2.08 % 4.98 % 2.56 % 100.00 % 2 s GPU @ 2.5 Ghz (Python)
54 PDSNet 2.29 % 4.05 % 2.58 % 100.00 % 0.5 s 1 core @ 2.5 Ghz (Python)
S. Tulyakov, A. Ivanov and F. Fleuret: Practical Deep Stereo (PDS): Toward applications-friendly deep stereo matching. Proceedings of the international conference on Neural Information Processing Systems (NIPS) 2018.
55 DeepStereo 2.16 % 4.72 % 2.59 % 100.00 % 0.9 s Titian X
56 SCV code 2.22 % 4.53 % 2.61 % 100.00 % 0.36 s Nvidia GTX 1080 Ti
C. Lu, H. Uchiyama, D. Thomas, A. Shimada and R. Taniguchi: Sparse Cost Volume for Efficient Stereo Matching. Remote Sensing 2018.
57 MRFnet 1.97 % 5.81 % 2.61 % 100.00 % 0.24 s GPU @ 2.5 Ghz (Python + C/C++)
58 DG-Net 2.06 % 5.47 % 2.63 % 100.00 % 0.03 s 1 core @ 2.5 Ghz (C/C++)
59 CooperativeStereo 2.09 % 5.38 % 2.64 % 100.00 % 0.9 s GPU @ 2.5 Ghz (Python + C/C++)
60 HTC
This method uses optical flow information.
 2.12 % 5.40 % 2.67 % 100.00 % 0.03 s 1 core @ 2.5 Ghz (C/C++)
61 CRL code 2.48 % 3.59 % 2.67 % 100.00 % 0.47 s Nvidia GTX 1080
J. Pang, W. Sun, J. Ren, C. Yang and Q. Yan: Cascade residual learning: A two-stage convolutional neural network for stereo matching. ICCV Workshop on Geometry Meets Deep Learning 2017.
62 CCFP-Net 2.11 % 5.53 % 2.68 % 100.00 % 0.5 s 8 cores @ 2.5 Ghz (Python)
63 NCCL2 2.11 % 5.59 % 2.69 % 100.00 % 0.61 s GPU @ 2.5 Ghz (Python + C/C++)
64 MFS-NET 2.22 % 5.09 % 2.70 % 100.00 % 0.5 s GPU @ 2.5 Ghz (Python)
65 ABN 2.20 % 5.35 % 2.73 % 100.00 % 0.08 s GPU @ 2.5 Ghz (Java)
66 CMF 2.22 % 5.41 % 2.75 % 100.00 % 0.5 s GPU @ 2.5 Ghz (Python)
ERROR: Wrong syntax in BIBTEX file.
67 GHSM-NET 2.48 % 4.29 % 2.78 % 100.00 % 0.2 s GPU @ 2.5 Ghz (Python)
68 GC-NET 2.21 % 6.16 % 2.87 % 100.00 % 0.9 s Nvidia GTX Titan X
A. Kendall, H. Martirosyan, S. Dasgupta, P. Henry, R. Kennedy, A. Bachrach and A. Bry: End-to-End Learning of Geometry and Context for Deep Stereo Regression. Proceedings of the International Conference on Computer Vision (ICCV) 2017.
69 CAR 2.35 % 5.53 % 2.88 % 100.00 % 0.5 s 1 core @ 2.5 Ghz (C/C++)
70 PSM-Cross 2.45 % 5.14 % 2.90 % 100.00 % 0.45 s GPU @ 2.5 Ghz (Python)
71 DPSM-Net 2.53 % 4.84 % 2.92 % 100.00 % 0.35 s GPU @ 2.5 Ghz (Python)
72 SemanStereo 2.36 % 5.72 % 2.92 % 100.00 % 60 s 1 core @ 2.5 Ghz (Python)
73 FBW_ROB 2.35 % 6.20 % 2.99 % 100.00 % 2 s GPU @ 2.5 Ghz (Python)
74 SPF-Net 2.60 % 4.97 % 2.99 % 100.00 % 0.16 s GPU @ 2.0 Ghz (Python + C/C++)
75 MCANet 2.82 % 3.90 % 3.00 % 100.00 % 0.33 s 1 core @ 2.5 Ghz (C/C++)
76 X_ASPP2 2.49 % 5.58 % 3.00 % 100.00 % 0.88 s GPU @ 2.5 Ghz (Python)
77 LRCR 2.55 % 5.42 % 3.03 % 100.00 % 49.2 s Nvidia GTX Titan X
Z. Jie, P. Wang, Y. Ling, B. Zhao, Y. Wei, J. Feng and W. Liu: Left-Right Comparative Recurrent Model for Stereo Matching. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2018.
78 CFCNet 2.47 % 5.90 % 3.04 % 100.00 % 0.47 s GPU @ 3.0 Ghz (Python)
79 anta-test-1 2.55 % 5.65 % 3.06 % 100.00 % 0.5 s GPU @ 1.5 Ghz (Python)
80 Fast DS-CS 2.83 % 4.31 % 3.08 % 100.00 % 0.02 s GPU @ 2.0 Ghz (Python + C/C++)
81 NVStereoNet code 2.62 % 5.69 % 3.13 % 100.00 % 0.6 s NVIDIA Titan Xp
N. Smolyanskiy, A. Kamenev and S. Birchfield: On the Importance of Stereo for Accurate Depth Estimation: An Efficient Semi-Supervised Deep Neural Network Approach. arXiv preprint arXiv:1803.09719 2018.
82 NVStereoNet_ROB 2.62 % 5.69 % 3.13 % 100.00 % 0.6 s NVIDIA Titan Xp
83 DRR 2.58 % 6.04 % 3.16 % 100.00 % 0.4 s Nvidia GTX Titan X
S. Gidaris and N. Komodakis: Detect, Replace, Refine: Deep Structured Prediction For Pixel Wise Labeling. arXiv preprint arXiv:1612.04770 2016.
84 MFMNet_s 2.97 % 4.20 % 3.17 % 100.00 % 0.36 s GPU @ 2.5 Ghz (Python + C/C++)
85 MFMNert 3.05 % 4.48 % 3.29 % 100.00 % 0.36 s GPU @ 2.5 Ghz (Python + C/C++)
86 CS2D 2.72 % 6.30 % 3.31 % 100.00 % 0.2 s 1 core @ 2.5 Ghz (Python)
87 SsSMnet 2.70 % 6.92 % 3.40 % 100.00 % 0.8 s P100
Y. Zhong, Y. Dai and H. Li: Self-Supervised Learning for Stereo Matching with Self-Improving Ability. arXiv:1709.00930 2017.
88 RTSnet 2.86 % 6.19 % 3.41 % 100.00 % 0.02 s P100 (pytorch)
89 L-ResMatch code 2.72 % 6.95 % 3.42 % 100.00 % 48 s 1 core @ 2.5 Ghz (C/C++)
A. Shaked and L. Wolf: Improved Stereo Matching with Constant Highway Networks and Reflective Loss. arXiv preprint arxiv:1701.00165 2016.
90 Displets v2 code 3.00 % 5.56 % 3.43 % 100.00 % 265 s >8 cores @ 3.0 Ghz (Matlab + C/C++)
F. Guney and A. Geiger: Displets: Resolving Stereo Ambiguities using Object Knowledge. Conference on Computer Vision and Pattern Recognition (CVPR) 2015.
91 anta-test-2 2.88 % 6.70 % 3.51 % 100.00 % 0.5 s GPU @ 1.5 Ghz (C/C++)
92 CNNF+SGM 2.78 % 7.69 % 3.60 % 100.00 % 71 s TESLA K40C
F. Zhang and B. Wah: Fundamental Principles on Learning New Features for Effective Dense Matching. IEEE Transactions on Image Processing 2018.
93 DH-SF
This method uses optical flow information.
 2.70 % 8.07 % 3.60 % 100.00 % 350 s 1 core @ 2.5 Ghz (Matlab + C/C++)
94 PBCP 2.58 % 8.74 % 3.61 % 100.00 % 68 s Nvidia GTX Titan X
A. Seki and M. Pollefeys: Patch Based Confidence Prediction for Dense Disparity Map. British Machine Vision Conference (BMVC) 2016.
95 SGM-Net 2.66 % 8.64 % 3.66 % 100.00 % 67 s Titan X
A. Seki and M. Pollefeys: SGM-Nets: Semi-Global Matching With Neural Networks. CVPR 2017.
96 Dense-CNN 2.90 % 8.79 % 3.88 % 100.00 % 53 s 1 core @ 2.5 Ghz (C/C++)
97 MC-CNN-acrt code 2.89 % 8.88 % 3.89 % 100.00 % 67 s Nvidia GTX Titan X (CUDA, Lua/Torch7)
J. Zbontar and Y. LeCun: Stereo Matching by Training a Convolutional Neural Network to Compare Image Patches. Submitted to JMLR .
98 ESM 3.33 % 6.73 % 3.90 % 100.00 % 0.03 s 1 core @ 2.5 Ghz (Python)
99 PASNet 3.28 % 7.45 % 3.97 % 100.00 % 0.6 s GPU @ 2.5 Ghz (Python)
100 RGL 4.22 % 4.02 % 4.19 % 100.00 % 0.1 s 1 core @ 2.5 Ghz (C/C++)
101 PRSM
This method uses optical flow information.
This method makes use of multiple (>2) views.
 code 3.02 % 10.52 % 4.27 % 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.
102 DispNetC code 4.32 % 4.41 % 4.34 % 100.00 % 0.06 s Nvidia GTX Titan X (Caffe)
N. Mayer, E. Ilg, P. Häusser, P. Fischer, D. Cremers, A. Dosovitskiy and T. Brox: A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation. CVPR 2016.
103 SGM-Forest 3.11 % 10.74 % 4.38 % 99.92 % 6 seconds 1 core @ 3.0 Ghz (Python/C/C++)
J. Schönberger, S. Sinha and M. Pollefeys: Learning to Fuse Proposals from Multiple Scanline Optimizations in Semi-Global Matching. European Conference on Computer Vision (ECCV) 2018.
104 SSF
This method uses optical flow information.
 3.55 % 8.75 % 4.42 % 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.
105 ISF
This method uses optical flow information.
 4.12 % 6.17 % 4.46 % 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.
106 MSFG-Net 3.62 % 8.90 % 4.50 % 100.00 % 0.6 s 1 core @ 2.5 Ghz (Python)
107 Content-CNN 3.73 % 8.58 % 4.54 % 100.00 % 1 s Nvidia GTX Titan X (Torch)
W. Luo, A. Schwing and R. Urtasun: Efficient Deep Learning for Stereo Matching. CVPR 2016.
108 MADnet 3.75 % 9.20 % 4.66 % 100.00 % 0.02 s GPU @ 2.5 Ghz (Python)
109 MSFG-Net 3.81 % 9.62 % 4.77 % 100.00 % 0.6 s GPU @ 2.5 Ghz (C/C++)
110 FastStereov2 3.91 % 9.19 % 4.79 % 100.00 % 0.03 s 1 core @ 2.5 Ghz (C/C++)
111 FastStereo 4.07 % 8.88 % 4.87 % 100.00 % 0.03 s 1 core @ 2.5 Ghz (C/C++)
112 MC-CNN-WS code 3.78 % 10.93 % 4.97 % 100.00 % 1.35 s 1 core 2.5 Ghz + K40 NVIDIA, Lua-Torch
S. Tulyakov, A. Ivanov and F. Fleuret: Weakly supervised learning of deep metrics for stereo reconstruction. ICCV 2017.
113 3DMST 3.36 % 13.03 % 4.97 % 100.00 % 93 s 1 core @ >3.5 Ghz (C/C++)
X. Lincheng Li and L. Zhang: 3D Cost Aggregation with Multiple Minimum Spanning Trees for Stereo Matching. submitted to Applied Optics .
114 SPOSF
This method uses optical flow information.
 4.12 % 9.49 % 5.01 % 99.96 % 10 min 1 core @ 3.5 Ghz (Matlab + C/C++)
115 OSF+TC
This method uses optical flow information.
This method makes use of multiple (>2) views.
 4.11 % 9.64 % 5.03 % 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.
116 CBMV code 4.17 % 9.53 % 5.06 % 100.00 % 250 s 6 cores @ 3.0 Ghz (Python,C/C++,CUDA Nvidia TitanX)
K. Batsos, C. Cai and P. Mordohai: CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation. 2018.
117 PWOC-3D
This method uses optical flow information.
 4.19 % 9.82 % 5.13 % 100.00 % 0.13 s GTX 1080 Ti
118 SS-SF
This method uses optical flow information.
 3.59 % 13.11 % 5.18 % 100.00 % 3 min 1 core @ 2.5 Ghz (Matlab + C/C++)
119 OSF 2018
This method uses optical flow information.
 code 4.11 % 11.12 % 5.28 % 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.
120 SPS-St code 3.84 % 12.67 % 5.31 % 100.00 % 2 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.
121 MDP
This method uses stereo information.
 4.19 % 11.25 % 5.36 % 100.00 % 11.4 s 4 cores @ 3.5 Ghz (Matlab + C/C++)
A. Li, D. Chen, Y. Liu and Z. Yuan: Coordinating Multiple Disparity Proposals for Stereo Computation. IEEE Conference on Computer Vision and Pattern Recognition 2016.
122 DC-NET 4.31 % 11.52 % 5.51 % 100.00 % 0.53 s >8 cores @ 3.5 Ghz (C/C++)
123 WDMC 4.64 % 10.33 % 5.59 % 100.00 % 1 min 8 cores @ 3.5 Ghz (Python)
124 SFF++
This method uses optical flow information.
This method makes use of multiple (>2) views.
 4.27 % 12.38 % 5.62 % 100.00 % 78 s 4 cores @ 3.5 Ghz (C/C++)
125 OSF
This method uses optical flow information.
 code 4.54 % 12.03 % 5.79 % 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.
126 VN 4.88 % 10.52 % 5.82 % 100.00 % 0.5 s GPU @ 3.5 Ghz (Python + C/C++)
127 SDR code 4.51 % 12.64 % 5.86 % 100.00 % 4.2 s 1 core @ 2.5 Ghz (C/C++)
128 pSGM 4.84 % 11.64 % 5.97 % 100.00 % 7.77 s 4 cores @ 3.5 Ghz (C/C++)
Y. Lee, M. Park, Y. Hwang, Y. Shin and C. Kyung: Memory-Efficient Parametric Semiglobal Matching. IEEE Signal Processing Letters 2018.
129 CSF
This method uses optical flow information.
 4.57 % 13.04 % 5.98 % 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.
130 MBM 4.69 % 13.05 % 6.08 % 100.00 % 0.13 s 1 core @ 3.0 Ghz (C/C++)
N. Einecke and J. Eggert: A Multi-Block-Matching Approach for Stereo. IV 2015.
131 PR-Sceneflow
This method uses optical flow information.
 code 4.74 % 13.74 % 6.24 % 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.
132 SGM+DAISY code 4.86 % 13.42 % 6.29 % 95.26 % 5 s 1 core @ 2.5 Ghz (C/C++)
133 DispSegNet 4.20 % 16.97 % 6.33 % 100.00 % 0.9 s GPU @ 2.5 Ghz (Python)
J. Zhang, K. Skinner, R. Vasudevan and M. Johnson-Roberson: DispSegNet: Leveraging Semantics for End- to-End Learning of Disparity Estimation from Stereo Imagery. arXiv preprint arXiv:1809.04734 2018.
134 DeepCostAggr code 5.34 % 11.35 % 6.34 % 99.98 % 0.03 s GPU @ 2.5 Ghz (C/C++)
A. Kuzmin, D. Mikushin and V. Lempitsky: End-to-end Learning of Cost-Volume Aggregation for Real-time Dense Stereo. 2017 IEEE 27th International Workshop on Machine Learning for Signal Processing (MLSP) 2017.
135 SGM_ROB 5.06 % 13.00 % 6.38 % 100.00 % 0.11 s Nvidia GTX 980
H. Hirschm\"uller: Stereo Processing by Semi-Global Matching and Mutual Information. IEEE Transactions on Pattern Analysis and Machine Intelligence 2008.
136 SceneFFields
This method uses optical flow information.
 5.12 % 13.83 % 6.57 % 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.
137 SPS+FF++
This method uses optical flow information.
 code 5.47 % 12.19 % 6.59 % 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.
138 UnOS(Full)
This method uses optical flow information.
 5.10 % 14.55 % 6.67 % 100.00 % 0.08 s 1 core @ 2.5 Ghz (C/C++)
139 FSF+MS
This method uses optical flow information.
This method makes use of the epipolar geometry.
This method makes use of multiple (>2) views.
 5.72 % 11.84 % 6.74 % 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.
140 AABM 4.88 % 16.07 % 6.74 % 100.00 % 0.08 s 1 core @ 3.0 Ghz (C/C++)
N. Einecke and J. Eggert: Stereo Image Warping for Improved Depth Estimation of Road Surfaces. IV 2013.
141 DLM-Net 5.04 % 15.76 % 6.83 % 100.00 % 0.68 s 1 core @ 2.5 Ghz (Python)
ERROR: Wrong syntax in BIBTEX file.
142 SGM+C+NL
This method uses optical flow information.
 code 5.15 % 15.29 % 6.84 % 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.
143 SGM+LDOF
This method uses optical flow information.
 code 5.15 % 15.29 % 6.84 % 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.
144 SGM+SF
This method uses optical flow information.
 5.15 % 15.29 % 6.84 % 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.
145 SMV 5.03 % 16.34 % 6.91 % 99.99 % 1.6 min 8 cores @ 3.5 Ghz (Python)
146 SNCC 5.36 % 16.05 % 7.14 % 100.00 % 0.08 s 1 core @ 3.0 Ghz (C/C++)
N. Einecke and J. Eggert: A Two-Stage Correlation Method for Stereoscopic Depth Estimation. DICTA 2010.
147 OASM-DDS 5.12 % 17.96 % 7.25 % 100.00 % 0.90 s 1 core @ 2.5 Ghz (Python)
148 DSimNet 6.15 % 13.20 % 7.32 % 100.00 % 0.57 s GPU @ 2.5 Ghz (Python)
149 WCMA_ROB 5.68 % 16.36 % 7.45 % 100.00 % 40 s 1 core @ 2.5 Ghz (Matlab + C/C++)
150 SGM+CT 6.50 % 16.62 % 8.18 % 99.53 % 23 s 1 core @ 2.5 Ghz (C/C++)
151 CSCT+SGM+MF 6.91 % 14.87 % 8.24 % 100.00 % 0.0064 s Nvidia GTX Titan X @ 1.0 Ghz (CUDA)
D. Hernandez-Juarez, A. Chacon, A. Espinosa, D. Vazquez, J. Moure and A. Lopez: Embedded real-time stereo estimation via Semi-Global Matching on the GPU. Procedia Computer Science 2016.
152 MeshStereo code 5.82 % 21.21 % 8.38 % 100.00 % 87 s 1 core @ 2.5 Ghz (C/C++)
C. Zhang, Z. Li, Y. Cheng, R. Cai, H. Chao and Y. Rui: MeshStereo: A Global Stereo Model With Mesh Alignment Regularization for View Interpolation. The IEEE International Conference on Computer Vision (ICCV) 2015.
153 PCOF + ACTF
This method uses optical flow information.
 6.31 % 19.24 % 8.46 % 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.
154 PCOF-LDOF
This method uses optical flow information.
 6.31 % 19.24 % 8.46 % 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.
155 OASM-Net 6.89 % 19.42 % 8.98 % 100.00 % 0.73 s GPU @ 2.5 Ghz (Python)
A. Li and Z. Yuan: Occlusion Aware Stereo Matching via Cooperative Unsupervised Learning. Proceedings of the Asian Conference on Computer Vision, ACCV 2018.
156 ELAS_ROB code 7.38 % 21.15 % 9.67 % 100.00 % 0.19 s 4 cores @ >3.5 Ghz (C/C++)
A. Geiger, M. Roser and R. Urtasun: Efficient Large-Scale Stereo Matching. ACCV 2010.
157 ELAS code 7.86 % 19.04 % 9.72 % 92.35 % 0.3 s 1 core @ 2.5 Ghz (C/C++)
A. Geiger, M. Roser and R. Urtasun: Efficient Large-Scale Stereo Matching. ACCV 2010.
158 REAF code 8.43 % 18.51 % 10.11 % 100.00 % 1.1 s 1 core @ 2.5 Ghz (C/C++)
C. Cigla: Recursive Edge-Aware Filters for Stereo Matching. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops 2015.
159 iGF
This method makes use of multiple (>2) views.
 8.64 % 21.85 % 10.84 % 100.00 % 220 s 1 core @ 3.0 Ghz (C/C++)
R. Hamzah, H. Ibrahim and A. Hassan: Stereo matching algorithm based on per pixel difference adjustment, iterative guided filter and graph segmentation. Journal of Visual Communication and Image Representation 2016.
160 OCV-SGBM code 8.92 % 20.59 % 10.86 % 90.41 % 1.1 s 1 core @ 2.5 Ghz (C/C++)
H. Hirschmueller: Stereo processing by semiglobal matching and mutual information. PAMI 2008.
161 SDM 9.41 % 24.75 % 11.96 % 62.56 % 1 min 1 core @ 2.5 Ghz (C/C++)
J. Kostkova: Stratified dense matching for stereopsis in complex scenes. BMVC 2003.
162 SGM&FlowFie+
This method uses optical flow information.
 11.93 % 20.57 % 13.37 % 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.
163 GCSF
This method uses optical flow information.
 code 11.64 % 27.11 % 14.21 % 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.
164 MT-TW-SMNet 15.47 % 16.25 % 15.60 % 100.00 % 0.4s GPU @ 2.5 Ghz (Python)
165 Mono-SF
This method uses optical flow information.
 14.21 % 26.94 % 16.32 % 100.00 % 41 s 1 core @ 3.5 Ghz (Matlab + C/C++)
166 CostFilter code 17.53 % 22.88 % 18.42 % 100.00 % 4 min 1 core @ 2.5 Ghz (Matlab)
C. Rhemann, A. Hosni, M. Bleyer, C. Rother and M. Gelautz: Fast Cost-Volume Filtering for Visual Correspondence and Beyond. CVPR 2011.
167 DWBSF
This method uses optical flow information.
 19.61 % 22.69 % 20.12 % 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.
168 monoResMatch 22.10 % 19.81 % 21.72 % 100.00 % 0.4 s GPU @ 2.5 Ghz (Python)
169 OCV-BM code 24.29 % 30.13 % 25.27 % 58.54 % 0.1 s 1 core @ 2.5 Ghz (C/C++)
G. Bradski: The OpenCV Library. Dr. Dobb's Journal of Software Tools 2000.
170 VSF
This method uses optical flow information.
 code 27.31 % 21.72 % 26.38 % 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.
171 SED code 25.01 % 40.43 % 27.58 % 4.02 % 0.68 s 1 core @ 2.0 Ghz (C/C++)
D. Pe\~{n}a and A. Sutherland: Disparity Estimation by Simultaneous Edge Drawing. Computer Vision -- ACCV 2016 Workshops: ACCV 2016 International Workshops, Taipei, Taiwan, November 20-24, 2016, Revised Selected Papers, Part II 2017.
172 DispCC 30.99 % 37.37 % 32.05 % 100.00 % 0.2 s 1 core @ 2.5 Ghz (C/C++)
173 MST code 45.83 % 38.22 % 44.57 % 100.00 % 7 s 1 core @ 2.5 Ghz (Matlab + C/C++)
Q. Yang: A Non-Local Cost Aggregation Method for Stereo Matching. CVPR 2012.
Table as LaTeX | Only published Methods




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 Stereo Evaluation: The classic stereo evaluation benchmark, featuring four test images in version 2 of the benchmark, with very accurate ground truth from a structured light system. 38 image pairs are provided in total.
  • Daimler Stereo Dataset: Stereo bad weather highway scenes with partial ground truth for freespace
  • Make3D Range Image Data: Images with small-resolution ground truth used to learn and evaluate depth from single monocular images.
  • Lubor Ladicky's Stereo Dataset: Stereo Images with manually labeled ground truth based on polygonal areas.
  • 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:
@ARTICLE{Menze2018JPRS,
  author = {Moritz Menze and Christian Heipke and Andreas Geiger},
  title = {Object Scene Flow},
  journal = {ISPRS Journal of Photogrammetry and Remote Sensing (JPRS)},
  year = {2018}
}
@INPROCEEDINGS{Menze2015ISA,
  author = {Moritz Menze and Christian Heipke and Andreas Geiger},
  title = {Joint 3D Estimation of Vehicles and Scene Flow},
  booktitle = {ISPRS Workshop on Image Sequence Analysis (ISA)},
  year = {2015}
}



eXTReMe Tracker