This table lists the benchmark results for the low-res two-view scenario. The benchmark evaluates the Middlebury stereo metrics:

The mask determines whether the metric is evaluated for all pixels with ground truth, or only for pixels which are visible in both images (non-occluded).
The coverage selector allows to limit the table to results for all pixels (dense), or a given minimum fraction of pixels.

Click one or more dataset result cells or column headers to show visualizations. Most visualizations are only available for training datasets. The visualizations may not work with mobile browsers.

Since we plan to add additional datasets soon (~ end of September), which will likely change the ranking, the average scores are currently still hidden.




Method Infoelect. 1lelect. 1select. 2lelect. 2select. 3lelect. 3sfacade 1lfacade 1sforest 1lforest 1sforest 2lforest 2sterra. 1lterra. 1sterra. 2lterra. 2s
sort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
MeshStereopermissivetwo views12.03 17.47 14.66 19.13 138.56 329.80 139.46 10.45 19.44 19.43 217.66 27.37 222.04 316.38 333.15 232.34 3
C. Zhang, Z. Li, Y. Cheng, R. Cai, H. Chao, Y. Rui: MeshStereo: A Global Stereo Model with Mesh Alignment Regularization for View Interpolation. ICCV 2015
ELAScopylefttwo views16.72 315.91 37.71 215.62 248.88 430.47 241.80 23.20 215.07 29.19 116.37 13.39 132.86 424.17 441.19 441.31 4
A. Geiger, M. Roser, R. Urtasun: Efficient large-scale stereo matching. ACCV 2010
SPS-STEREOcopylefttwo views16.06 219.74 419.91 420.06 433.66 237.68 345.60 37.23 319.08 313.17 435.07 416.66 416.56 115.26 229.55 124.45 1
K. Yamaguchi, D. McAllester, R. Urtasun: Efficient Joint Segmentation, Occlusion Labeling, Stereo and Flow Estimation. ECCV 2014
SGM-STEREOtwo views17.89 412.77 218.42 319.55 330.39 140.48 446.73 411.67 419.99 412.56 333.65 315.10 317.44 213.30 133.35 328.26 2