We fixed a rounding issue on 1/8/2017. This affected the error numbers for the three "old" image pairs with integer ground truth (test/Computer, training/ArtL, and training/Teddy), for which submitted disparities should have been rounded to integers (at full resolution) before evaluation.
OpenCV 2.4.8 StereoSGBM method, full variant (2 passes). Reimplementation of H. Hirschmüller's SGM method (CVPR 2006; PAMI 2008).
OpenCV's "semi-global block matching" method; memory-intensive 2-pass version, which can only handle the quarter-size images. The matching cost is the sum of absolute differences over small windows. Aggregation is performed by dynamic programming along paths in 8 directions. Post filter as implemented in OpenCV. Dense results are created by hole-filling along scanlines.
SAD window: 3x3 pixel
Truncation value for pre-filter: 63
P1/P2: 8*3*3*3/32*3*3*3
Uniqueness ratio: 10
Speckle window size: 100
Speckle range: 32
Full DP: true
C/C++; 1 core, i7@3.3 GHz
07/25/14
1
SGBM2
Q
1
26.6
36
27.9
28
12.1
38
17.8
42
13.7
24
74.5
46
16.2
34
30.3
32
26.3
37
11.0
40
64.4
47
37.9
39
25.8
33
25.3
34
29.3
39
43.7
38
OpenCV 2.4.8 StereoSGBM method, single-pass variant. Reimplementation and modification of H. Hirschmüller's SGM method (CVPR 2006; PAMI 2008).
OpenCV's "semi-global block matching" method; memory efficient single-pass version. The matching cost is the sum of absolute differences over small windows. Aggregation is performed by dynamic programming along paths in only 5 of 8 directions. Post filter as implemented in OpenCV. Dense results are created by hole-filling along scanlines.
SAD window: 3x3 pixel
Truncation value for pre-filter: 63
P1/P2: 8*3*3*3/32*3*3*3
Uniqueness ratio: 10
Speckle window size: 100
Speckle range: 32
Full DP: false
The images are Census transformed and the Hamming distance is used as pixelwise matching cost. Aggregation is performed by a kind of dynamic programming along 8 paths that go from all directions through the image. Small disparity patches are invalidated. Interpolation is also performed along 8 paths.
OpenCV 2.4.8 StereoSGBM method, single-pass variant. Reimplementation and modification of H. Hirschmüller's SGM method (CVPR 2006; PAMI 2008).
OpenCV's "semi-global block matching" method; memory efficient single-pass version. The matching cost is the sum of absolute differences over small windows. Aggregation is performed by dynamic programming along paths in only 5 of 8 directions. Post filter as implemented in OpenCV. Dense results are created by hole-filling along scanlines.
SAD window: 3x3 pixel
Truncation value for pre-filter: 63
P1/P2: 8*3*3*3/32*3*3*3
Uniqueness ratio: 10
Speckle window size: 100
Speckle range: 32
Full DP: false
C/C++; 1 core, i7@3.3 GHz
07/25/14
4
SGBM1
F
3
28.4
40
43.5
39
9.09
31
13.6
34
25.9
39
82.0
49
14.4
27
43.4
40
30.3
41
5.98
27
59.3
46
45.8
44
28.5
39
24.9
32
20.1
32
45.9
40
OpenCV 2.4.8 StereoSGBM method, single-pass variant. Reimplementation and modification of H. Hirschmüller's SGM method (CVPR 2006; PAMI 2008).
OpenCV's "semi-global block matching" method; memory efficient single-pass version. The matching cost is the sum of absolute differences over small windows. Aggregation is performed by dynamic programming along paths in only 5 of 8 directions. Post filter as implemented in OpenCV. Dense results are created by hole-filling along scanlines.
SAD window: 3x3 pixel
Truncation value for pre-filter: 63
P1/P2: 8*3*3*3/32*3*3*3
Uniqueness ratio: 10
Speckle window size: 100
Speckle range: 32
Full DP: false
The images are Census transformed and the Hamming distance is used as pixelwise matching cost. Aggregation is performed by a kind of dynamic programming along 8 paths that go from all directions through the image. Small disparity patches are invalidated. Interpolation is also performed along 8 paths.
The images are Census transformed and the Hamming distance is used as pixelwise matching cost. Aggregation is performed by a kind of dynamic programming along 8 paths that go from all directions through the image. Small disparity patches are invalidated. Interpolation is also performed along 8 paths.
Correlation with five, partly overlapping windows on Census transformed images using Hamming distance as matching cost. A left-right consistency check ensures unique matches and filtering small disparity segments removes outliers. Interpolation is done within image rows with the lowest, valid neighboring disparity.
Census window: 7x7 pixel
Correlation window: 9x9 pixel
LR-check: on
Min. segments: 200 pixel
Interpolation: horizontal, lowest neighbor
A fast method for high-resolution stereo matching without exploring the full search space. Plane hypotheses are generated from sparse feature matches. Around each plane, a local plane sweep with +/- 3 disparities levels is performed to establish local disparity hypotheses via SGM using NCC matching costs. Finally, each pixel is assigned to one hypothesis using global optimization, again using SGM.
nRounds=3
The full set of parameters is listed in the paper and the supplemental materials on the project webpage.
A fast method for high-resolution stereo matching without exploring the full search space. Plane hypotheses are generated from sparse feature matches. Around each plane, a local plane sweep with +/- 3 disparities levels is performed to establish local disparity hypotheses via SGM using NCC matching costs. Finally, each pixel is assigned to one hypothesis using global optimization, again using SGM.
nRounds=3
The full set of parameters is listed in the paper and the supplemental materials on the project webpage.
This approach is an adaptive local stereo-method. It is integrated into a hierarchical scheme, which exploits adaptive windows. Sub-pix disparities are estimated,but not refined.
L = 10
t = 35
medianK = [3 3]
censusK = [9 7]
lambda = 45;
Block-matching stereo with Summed Normalized Cross-Correlation (SNCC) measure. Standard post-processed is applied, including a left-right check, error island removal (region growing), hole-filling and median filtering.
SNCC (first stage 3x3, second stage 11x11)
min correlation threshold = 0.3
region growing threshold = 2.5 disparity
min region size = 200 pixel
median filter = 1x5 and 5x1
Efficient two-pass aggregation with census/gradient cost metric, followed by iterative cost penalization and disparity re-selection to encourage local smoothness of disparities.
census window size = 9 x 7
max census distance = 38.03
max gradient difference = 2.51
census/gradient balance = 0.09
aggregation window size = 33 x 33
aggregation range parameter = 23.39
aggregation spatial parameter = 7.69
refinement window size = 65 x 65
refinement range parameter = 11.30
refinement spatial parameter = 17.20
cost penalty coefficient = 0.0023
median filter window size = 3 x 3
3 iterations of refinement
confidence threshold of 0.1 for sparse maps
CUDA C++, NVIDIA GeForce TITAN Black
10/07/14
14
IDR
H
2
18.4
23
37.5
35
4.08
6
7.49
13
23.3
35
40.6
28
15.7
33
24.5
26
11.3
22
5.46
23
33.1
26
26.0
21
21.5
25
21.7
26
15.3
22
21.2
20
Anonymous. Using local cues to improve dense stereo matching. CVPR 2015 submission 973.
In stereo matching cost filtering methods and energy minimization algorithms are considered as two different techniques. Due to their global extend energy minimization methods obtain good stereo matching results. However, they tend to fail in occluded regions, in which cost filtering approaches obtain better results. In this paper we intend to combine both approaches with the aim to improve overall stereo matching results.
We propose to perform stereo matching as a two-step energy minimization algorithm. We consider two MRF models: a fully connected model defined on the complete set of pixels in an image and a conventional locally connected model. We solve the energy minimization problem for the fully connected model, after which the marginal function of the solution is used as the unary potential in the locally connected MRF model.
Only gradient component (6D vector) of color images is used
A local matching technique utilizing SAD+Census cost measure and a recursive edge-aware aggregation through Successive Weighted Summation. Occlusion handling is provided via left-right cross check and a background favored filling.
smoothness parameter sigma = 24
5x5 Census window, Census weight=0.7, SAD weight=0.3, occlusion threshold=2
This approach triangulates the polygonized SLIC segmentations of the input images and optimizes a lower-layer MRF on the resulting set of triangles defined by photo consistency and normal smoothness. The lower-layer MRF is solved by a quadratic relaxation method which iterates between PatchMatch and Cholesky Decomposition. The lower-layer MRF is assisted by a upper-layer MRF defined on the set of triangle vertices which exploits local 'visual complexity' cues and encourages smoothness of the vertices' splitting properties. The two layers interact through an Alignment energy term which requires triangles sharing a non-split vertex to have their disparities agree on that vertex. Optimization of the whole model is iterated between optimizations of the two layers till convergence where the upper-layer can be solved in closed form.
omega=0.2
tau_grad=15
theta goes from 0 to 100 by smoothstep function in ten iterations
gamma1=30
gamma2=60
gamma3=0.8
Compute the matching cost with a convolutional neural network (accurate architecture). Then apply cross-based cost aggregation, semiglobal matching, left-right consistency check, median filter, and a bilateral filter.
DETAILS:
The network is similar to the one described in our CVPR paper differing only in the values of some hyperparameters. The input to the network are two 11 x 11 image patches. Five convolutional layers with 3 x 3 kernels and 112 feature maps extract feature vectors from the input image patches. The two 112-length feature vectors are concatenated into a 224-length vector which is passed through three fully-connected layers with 384 units each. The final (fourth) fully-connected layer projects the output to a single number---the matching cost. One important addition was the use of data augmentation techniques to increase the size of the training set. We tried to use as much training data as possible. Therefore we combined all of the 2001, 2003, 2005, 2006, and 2014 Middlebury datasets obtaining 60 image pairs. For the newer datasets (2005, 2006, and 2014) we also used several illumination and exposure settings.
A prior disparity image is calculated by matching a set of reliable support points and triangulating between them. A maximum a-posterior approach refines the disparities. The disparities for the left and right image are checked for consistency and disparity segments below a size of 50 pixels removed. (Improved results as of 9/14/2015 due to bug fix in color-to-gray conversion.)
Standard parameters of Libelas as provided with the MiddEval3-SDK.
A prior disparity image is calculated by matching a set of reliable support points and triangulating between them. A maximum a-posterior approach refines the disparities. The disparities for the left and right image are checked for consistency and disparity segments below a size of 50 pixels removed. (Improved results as of 9/14/2015 due to bug fix in color-to-gray conversion.)
Standard parameters of Libelas as provided with the MiddEval3-SDK.
C++/SSE; 1 core, i7@3.6 GHz
09/14/15
22
ELAS
H
2
27.3
38
43.3
38
12.5
40
13.9
38
23.7
36
66.1
41
20.4
38
33.1
34
20.5
32
11.0
39
43.9
37
37.8
38
26.4
34
28.6
39
38.3
43
33.3
33
S. Fang and Y. Li. Removed based multi-view stereo using window-based matching method. Submitted to MV&A, 2015.
Removed erroneous corresponding points from stereo pair and only correct corresponding points are kept which are obtained by NCC.
The method generates multiple proposals on absolute and relative disparities from multi-segmentations. The proposals are coordinated by point-wise competition and pairwise collaboration within a MRF model. During inference, a dynamic programming is performed in different directions with various step sizes.
We post-process the depth maps produced by Zbontar & LeCun's MC-CNN technique. We use a domain transform to compute an edge-aware variance measure of our confidence in the depth map, and then run our robust bilateral solver on that depth map and confidence with a Geman-McClure loss function.
The MC-CNN is computed using the publicly-available implementation (https://github.com/jzbontar/mc-cnn) which using the GPU, and the robust bilateral solver is computed using our CPU implementation which does not use the GPU, and is written in vanilla C++.
Intel(R) Xeon(R) CPU E5-1650 0 @ 3.20GHz, 6 cores; 32 GB RAM; NVIDIA GTX TITAN X
11/03/15
26
MC-CNN+RBS
H
2
8.62
7
6.05
10
5.16
20
6.24
8
3.27
6
11.1
8
8.91
7
8.87
11
9.83
18
3.21
9
15.1
8
15.9
9
12.8
6
13.5
4
7.04
7
9.99
3
Anonymous. High accuracy stereo matching with spatially varying regularization. CVPR 2016 submission 863.
C++; 1 i7 core @2.8GHz + Nvidia GTX 660Ti GPU
11/05/15
27
GCSVR
H
2
14.8
19
17.1
20
3.50
3
8.22
17
16.5
28
47.4
33
11.4
16
9.75
14
7.06
8
3.17
8
34.4
28
27.1
22
18.3
17
19.2
19
16.0
24
19.3
18
Anonymous. Image-guided non-local dense matching with three-steps optimization. ISPRS Congress 2016 submission 231.
This paper proposes a new image-guided non-local dense matching method with a three-step optimization based on the combination of image-guided methods and energy function-guided methods.
Cost Computation:
Window Size: 5
Weighting Coefficient: 0.3
Truncation Threshold (Census): 15
Truncation Threshold (HOG): 1
Image-guided Non-local Matching:
Smooth Term: 6
Penalty Term P1: 0.3
Penalty Term P2: 3
Disparity Interpolation:
Truncation Threshold: 5
Smooth Term: 3
Penalty Term P1: 3
Penalty Term P2: 30
Function Base: 5
An efficient stereo matching algorithm, which applies adaptive smoothness constraints using texture and edge information, is proposed in this work. First, we determine non-textured regions, on which an input image yields flat pixel values. In the non-textured regions, we penalize depth discontinuity and complement the primary CNN-based matching cost with a color-based cost. Second, by combining two edge maps from the input image and a pre-estimated disparity map, we extract denoised edges that correspond to depth discontinuity with high probabilities. Thus, near the denoised edges, we penalize small differences of neighboring disparities.
The method uses the MC-CNN code for the matching cost computation only.
Compute the matching cost with a convolutional neural network (fast architecture). Then apply cross-based cost aggregation, semiglobal matching, left-right consistency check, median filter, and a bilateral filter.
See paper.
NVIDIA GTX TITAN X
01/26/16
32
MC-CNN-fst
H
2
9.69
10
7.35
13
5.07
19
7.18
12
4.71
9
16.8
14
11.2
14
7.37
7
6.97
7
2.82
5
20.7
12
17.4
12
15.4
12
15.1
9
7.90
9
12.6
11
Anonymous. Stereo depth map refinement with scene layout estimation. CVPR 2016 submission 617.
We exploit scene layout information to refine depth maps.
Anonymous. LS-ELAS: line segment based efficient large scale stereo matching. ICRA 2017 submission 289.
Our approach is an extension of the ELAS (from Geiger et al.) algorithm. We extract edges and sample our candidate support points along them. For every two consecutive valid support points we create a (straight) line segment. We force the triangulation to include the set of line segments (constrained Delaunay) for a better preservation of the disparity discontinuity at the edges.
Parameters as in the original ELAS algorithm.
For sampling candidate support points along the edge segments:
Adaptive sampling activated:
step = ceil(sqrt(img_diag)*0.5);
sampler(sqrt(step) / 2, step / 2, step / 2);
Anonymous. Morphological processing of stereoscopic image superimpositions for disparity map estimation. ECCV 2016 submission 1308.
The computation of the sparse disparity maps is achieved by means of a 3D diffusion of the costs contained in the disparity space volume. The watershed segmentations of the left and right views control the diffusion process and valid measurements are obtained by cross-checking.
The estimation of the dense disparity maps uses the sparse measurements as control points and is driven by a 3D watershed separating the disparity space volume into foreground and background pixels.
Python; 1 i5 core @2.7Ghz
03/15/16
34
MPSV
Q
1
43.5
49
58.8
49
33.9
53
34.2
51
37.9
46
52.4
35
30.8
46
56.8
49
51.0
50
30.6
50
56.9
42
51.5
48
44.6
49
43.4
49
44.2
46
54.2
44
Shahbazi et al. Revisiting intrinsic curves for efficient dense stereo matching. ISPRS Congress 2016 submission 913.
No post processing (no filtering, no hole-filling, no interpolation) performed.
The concepts of intrinsic curves were revisited and used for:
- disparity search space reduction, resulting in 83% reduction of the disparity range (individually for each pixel) directly from the original resolution of the image without needing hierarchical search
- reducing the ambiguities due to occluded pixels by integrating occlusion clues explicitly into the global energy function as a soft prior
The final energy minimization was done using semi global approach along eight paths.
Matching (data) cost = census transform 7*9
Occlusion cost= from intrinsic curves curvature
C++; 1 i7-2760QM CPU @ 2.4 GHz
04/03/16
35
ICSG
F
3
45.6
50
69.7
53
19.1
45
21.3
46
43.6
49
77.6
47
36.9
50
65.3
50
40.4
46
20.3
45
53.6
40
58.7
50
46.5
50
47.1
51
60.7
49
79.1
50
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. Submitted to IJCV 2016.
An extended version of MeshStereo model. Use matching cost (matching cost only) computed from MC-CNN of Zbontar and LeCun.
See paper of MeshStereo, ICCV 2015
C/C++, 8 cores + NVIDIA TITAN X
04/12/16
36
MeshStereoExt
H
2
7.29
3
4.41
2
3.98
5
5.40
4
3.17
5
10.0
4
8.89
6
4.62
4
4.77
5
3.49
11
12.7
3
12.4
3
10.4
2
14.5
6
7.80
8
8.85
2
Z. Ge. A global stereo matching algorithm with iterative optimization. China CAD & CG 2016 submission 595.
the basic idea is treat stereo matching as a Markov Random Field problem,and iterative optimize the initial result use modified EM-like algorithm
S. Hadfield, K. Lebeda, and R. Bowden. Stereo reconstruction using top-down cues. Submitted to CVIU 2016.
Incorporating cues from top-down (holistic) scene understanding into existing bottom-up stereo reconstruction techniques (CoR - Chakrabarti et al. CVPR 2015).
Learned weightings (from 2006 dataset) for High Level Scene Cues. Default parameters for CoR. Images with max disp > 256 were downsampled before the SGM step of CoR.
Matlab and C++. single E5 core at 2.4GHz
04/24/16
38
HLSC_cor
H
2
26.4
35
26.5
27
15.2
42
21.0
45
20.5
33
35.7
24
27.7
44
33.1
35
35.0
44
11.9
42
39.1
32
34.2
29
25.2
31
32.8
44
28.3
38
22.7
21
Anonymous. Stereo matching by joint energy minimization. ECCV 2016 submission 41.
1 i7 Core @3.5GHz
04/27/16
39
JEM
Q
1
37.2
45
35.7
34
27.9
49
30.6
50
33.2
43
43.0
29
31.4
48
49.5
46
47.3
47
26.5
48
49.6
39
46.0
46
35.7
45
30.8
42
37.5
42
55.8
45
L. Li, S. Zhang, X. Yu, and L. Zhang. PMSC: PatchMatch-based superpixel cut for accurate stereo matching. Submitted to IEEE Transactions on Circuits and Systems for Video Technology, 2016.
A 3D label based method with global optimization at pixel level. A bilayer matching cost is employed by first matching small square windows then aggregate on large irregular windows. Global optimization is carried out by fusing candidate proposals, which are generated from our specific superpixel structure.
M.-G. Park and K.-J. Yoon. As-planar-as-possible depth map estimation. Submitted to IEEE TPAMI 2016.
We exploit local and dominant plane hypotheses to estimate as-planar-as-possible disparity maps.
beta = 1,000, lambda_aps = 28
lambda_per = 10
Matlab+Mex; i7 core 3.5GHz; 4 cores
05/28/16
41
APAP-Stereo
H
2
7.46
4
5.43
4
4.91
16
5.11
2
5.17
10
21.6
16
9.50
8
4.31
2
4.23
4
3.24
10
14.3
7
9.78
1
7.32
1
13.4
3
6.30
4
8.46
1
Anonymous. Look wider and deeper to match. CVPR 2016 submission 975.
.
.
i7 core @3.2GHz + GTX 980 GPU
11/06/15
28
SOU4P-net
H
2
13.5
17
23.1
24
5.41
22
6.39
9
13.1
23
30.5
21
11.1
13
16.4
21
12.7
24
3.13
7
28.9
17
17.1
11
16.4
14
16.9
12
10.7
15
14.5
12
Anonymous. 3D labeling stereo matching with content aware adaptive windows. 3DV 2016 submission 25.
Matlab, core i5, 4 cores+ 2xGTX 970
07/03/16
42
LPU
H
2
10.5
11
11.4
15
3.18
2
8.10
15
6.08
12
20.9
15
9.84
9
6.94
6
4.00
1
4.04
14
33.9
27
16.9
10
15.2
11
17.8
15
9.12
10
11.6
8
Anonymous. Disparity estimation by simultaneous edge drawing. ACCV 2016 Workshop 1 - 3D modelling and applications - Submission id 18.
CRT edge point matching + Edge based disparity propagation
TBD
C++; 1 i7 core @ 2.1Ghz
08/31/16
43
SED
F
3
63.4
52
54.3
46
22.4
47
72.9
54
64.5
53
71.4
44
42.5
53
80.1
52
67.9
52
49.8
54
79.6
50
74.4
52
65.4
54
55.1
52
86.1
54
91.6
54
Anonymous. Robust stereo matching with surface normal prediction. ICRA 2017 submission 440.
Matlab; 1 i5-4590 CPU@3.3 GHz
09/13/16
44
SNP-RSM
H
2
8.98
8
5.46
5
4.85
14
6.50
10
3.37
7
10.4
6
10.1
11
8.73
10
9.37
16
3.58
12
14.3
6
14.7
8
14.9
9
12.8
1
10.1
13
10.8
6
Anonymous. A learned sparseness and IGMRF based regularization framework for dense disparity estimation using unsupervised feature learning. Submitted to IPSJ Transactions on Computer Vision and Applications, 2016.
Dense disparity estimation in a sparsity and IGMRF based regularization framework where the matching is performed using learned features and intensities of stereo images.
manually set
Core i7-3632QM, 2.20 GHz processor and 8.00 GB RAM.
09/20/16
45
LFSIR
Q
1
70.1
54
75.7
54
60.3
54
67.1
52
72.4
54
80.8
48
53.7
54
85.4
54
83.8
54
42.5
53
91.2
52
90.4
53
64.1
53
71.3
54
61.5
50
90.3
53
H. Park and K. Lee. Look wider to match image patches with convolutional neural network. Submitted to IEEE Signal Processing Letters, 2016.
A novel pooling scheme is used to train a matching cost function with a CNN. It widens the size of receptive field effectively without losing the fine details.
The overall post-processing pipeline is kept almost same as the original MC-CNN-acrt, except that the parameter settings are changed as follows:
cbca_num_iterations_1 = 0, cbca_num_iterations_2 = 1, sgm_P1 = 1.3, sgm_P2 = 17.0, sgm_Q1 = 3.6, sgm_Q2 = 36.0, and sgm_V = 1.4.
Torch; the Intel core i7 4790K
CPU and a single Nvidia Geforce GTX Titan X GPU
10/19/16
46
LW-CNN
H
2
7.23
2
4.65
3
3.95
4
5.30
3
2.63
2
11.2
9
7.86
2
4.32
3
4.22
3
2.43
3
12.2
2
13.4
4
13.6
8
14.8
8
4.72
1
12.0
9
M. Joshi. A learned IGMRF sparseness and IGMRF based regularization framework for dense disparity estimation. Submitted to IPSJ CVA 2016.
An energy minimization framework for disparity estimation where energy function consists of intensity matching cost, feature matching cost, IGMRF prior and sparsity priors.
Manually set
MATLAB 2014 @2.22 Ghz
10/23/16
47
SIGMRF
Q
1
64.2
53
60.0
51
33.0
52
67.9
53
63.2
52
99.5
54
39.8
51
84.8
53
82.0
53
35.2
52
95.2
54
91.5
54
58.1
52
65.8
53
55.0
48
88.6
52
Anonymous. High-resolution stereo matching based on sampled photoconsistency computation. CVPR 2017 submission 1351.
C/C++ 6 cores Intel Core i7-5820K @3.3 GHz
11/06/16
48
SPS
F
3
19.6
26
14.2
17
12.3
39
14.9
40
12.0
20
15.8
12
19.1
37
17.4
22
15.4
25
8.23
33
30.9
21
34.8
32
30.6
41
25.3
34
28.3
37
28.0
28
Anonymous. End-to-end training of hybrid CNN-CRF models for fast and accurate stereo. CVPR 2017 submission 853.
NVidia TitanX
11/11/16
49
JMR
H
2
16.5
20
27.9
29
4.79
13
8.62
18
20.2
32
37.6
26
15.1
31
24.2
25
17.4
29
4.47
17
19.2
10
22.0
16
20.0
19
21.1
24
9.43
11
16.2
14
Anonymous. Is domain expertise obsolete in stereo matching? CVPR 2017 submission 3635.
Anonymous. Semi-supervised learning of deep metrics for stereo reconstruction. CVPR 2017 submission 1205.
This is a new semi-supervised method that allows to learn deep metrics from unlabeled stereo images, given coarse information about the scenes and the optical system. The network architecture is similar to MC-CNN-fst.
1 core 2.5 Ghz + K40 NVIDIA, Lua-Torch
11/15/16
51
MC-CNN-SS
H
2
12.3
14
14.8
19
7.20
27
11.1
28
7.62
14
15.9
13
14.3
26
11.5
17
9.01
14
3.89
13
19.7
11
20.5
14
16.3
13
16.3
11
12.1
17
18.3
17
Anonymous. Learning to compute the stereo matching cost without supervision. CVPR 2017 submission 2151.
SGM type method where the matching cost is computed with a CNN trained without ground truth data.
See paper
NVIDIA Titan X
11/15/16
52
UCNN
H
2
20.5
29
44.8
40
9.77
34
13.6
34
18.2
29
36.5
25
12.8
17
23.4
24
12.4
23
9.22
37
39.5
34
30.5
26
24.8
30
21.2
25
19.1
29
32.3
32
Anonymous. Simultaneous learning matching cost and smoothness constraint for stereo matching. CVPR 2017 submission 1368.
C/C++; 4 cores + GTX 1080 GPU
11/16/16
53
MCSC
F
3
11.3
13
13.3
16
5.96
24
10.6
24
8.69
15
7.22
2
11.3
15
10.6
15
7.48
9
3.07
6
3.10
1
25.2
20
19.0
18
17.2
14
10.3
14
25.5
26
N. Ma, Y. Men, C. Men, and X. Li. Accurate dense stereo matching based on image segmentation using an adaptive multi-cost approach. Submitted to Symmetry, 2016.
This is a segmentation based stereo matching algorithm using an adaptive multi-cost approach, which is exploited for obtaining accuracy disparity maps.