Stereo Dataset
6 stereo-pairs with groundtruth
This page provides a stereo dataset composed of 6 stereo-pairs with
groundtruth acquired by Federico Tombari and Andrea Mainetti. The groundtruth was obtained by means of an active stereo technique (space-time stereo):
details concerning the setup used for the acquisition can be found in
[1]. In
addition to groundtruth, each stereo pair comes with a binary mask
that indicate
erroneous pixels in the provided ground-truth mask: this can be useful,
e.g., to evaluate quantitatively a disparity map by comparing it with
the provided groundtruth.
This dataset was presented
in the stereo evaluation proposed in [2]. If you use this dataset for
scientific or research purposes, please cite appropriately this webpage
and [2]. For any question feel free to write me at: federico (DOT) tombari (AT) unibo (DOT) it
DOWNLOAD
The dataset can be downloaded as a unique zip file from this link:
Dataset (zip format)
EXAMPLE
Below we show one of the 6 stereo-pairs of the dataset including Left Image, Right Image, ground-truth and binary mask.
Left Image |
|
Right
Image |
|
Ground Truth |
|
Mask |
|
REFERENCES
[1] F.
Tombari, L. Di Stefano, S. Mattoccia, A.Mainetti, "A 3D reconstruction system based on improved Spacetime Stereo", 11th Int. Conf. on Control, Automation, Robotics and Vision (ICARCV 2010), December 7-10, 2010, Singapore
[2] F.
Tombari, S. Mattoccia, L. Di Stefano, "Stereo for robots: quantitative evaluation of efficient and low-memory dense stereo algorithms", 11th Int. Conf. on Control, Automation, Robotics and Vision (ICARCV 2010), December 7-10, 2010, Singapore
|