![]() |
CMP Stereo Algorithm Evaluation |
Center for Machine Perception Czech Technical University Prague http://cmp.felk.cvut.cz/ |
Jana Kostliva Radim Sara Jan Cech |
Introduction | Results | Download | Upload |
This website presents material for stereo matching (dense or semi-dense) algorithm evaluation described in [1,2]. It contains matlab toolbox for evaluating your algorithm, comparison of already tested algorithms, downloads with scenes or results and upload form for uploading your results.
Our methodology evaluates algorithms over a wide range of parameter settings which allows to get the algorithm performance as good as possible. The results are presented as error-plots as well as numerical characteristics describing the algorithm performance.
Our evaluation uses also scenes from Middlebury dataset (Cones, Teddy, Tsukuba, and Venus), courtesy of D. Scharstein, which has been introduced in [3,4].
We invite other researches to contribute their results to this page to create wider a study. To this end, we provide a matlab toolbox which helps you to run the evaluation on your algorithm. If you are interested in contributing, you can download it on page "Downloads".
We give permission to use and publish all images and results on this website. Nevertheless, if you use our toolbox and/or results created based on this method, please cite the appropriate articles [1,2].
[1] |
J. Kostlivá, J.Čech, R.©ára: Feasibility Boundary in Dense and
Semi-Dense Stereo Matching. In BenCOS2007, in conjunction with
CVPR2007, on-line, 2007. Best Paper Award
|
[2] |
J. Kostlivá, J.Cech, R.©ára: ROC Based Evaluation of Stereo
Algorithms. Research report CTU-CMP-2007-08, Center for Machine
Perception, K13133 FEE, Czech Technical University, Prague, Czech
Republic, 2007.
|
[3] | D. Scharstein, R. Szeliski: A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms. IJCV 47(1/2/3):7-42, 2002. |
[4] |
D. Scharstein, R. Szeliski: High-accuracy Stereo Depth Maps Using
Structured Light. In CVPR 2003, pages 195-202, 2002.
|
Go to the cmp homepage! | comments and suggestions to Authors |