We compare two approaches to the problem of estimating the depth of a point in space from observing its image position in two different cameras: 1.~The classical photogrammetric approach explicitly models the two cameras and estimates their intrinsic and extrinsic parameters using a tedious calibration procedure; 2.~A generic machine learning approach where the mapping from image to spatial coordinates is directly approximated by a Gaussian Process regression. Our results show that the generic learning approach, in addition to simplifying the procedure of calibration, can lead to higher depth accuracies than classical calibration although no specific domain knowledge is used.
| Author(s): | Sinz, F. and Candela, JQ. and BakIr, G. and Rasmussen, CE. and Franz, M. |
| Book Title: | 26th DAGM Symposium |
| Journal: | Pattern Recognition: 26th DAGM Symposium |
| Pages: | 245-252 |
| Year: | 2004 |
| Month: | September |
| Day: | 0 |
| Series: | LNCS 3175 |
| Editors: | Rasmussen, C. E., H. H. B{\"u}lthoff, B. Sch{\"o}lkopf, M. A. Giese |
| Publisher: | Springer |
| Bibtex Type: | Conference Paper (inproceedings) |
| Address: | Berlin, Germany |
| Event Name: | 26th DAGM Symposium |
| Event Place: | Tübingen, Germany |
| Digital: | 0 |
| Electronic Archiving: | grant_archive |
| Institution: | Deutsche Arbeitsgemeinschaft für Mustererkennung e.V. |
| Language: | en |
| Organization: | Max-Planck-Gesellschaft |
| School: | Biologische Kybernetik |
| Links: | |
BibTex
@inproceedings{2644,
title = {Learning Depth From Stereo},
journal = {Pattern Recognition: 26th DAGM Symposium},
booktitle = {26th DAGM Symposium},
abstract = {We compare two approaches to the problem of estimating the depth
of a point in space from observing its image position in two
different cameras: 1.~The classical photogrammetric approach
explicitly models the two cameras and estimates their intrinsic
and extrinsic parameters using a tedious calibration procedure;
2.~A generic machine learning approach where the mapping from
image to spatial coordinates is directly approximated by a Gaussian Process regression. Our results show that the generic
learning approach, in addition to simplifying the procedure of
calibration, can lead to higher depth accuracies than classical
calibration although no specific domain knowledge is used.},
pages = {245-252},
series = {LNCS 3175},
editors = {Rasmussen, C. E., H. H. B{\"u}lthoff, B. Sch{\"o}lkopf, M. A. Giese},
publisher = {Springer},
organization = {Max-Planck-Gesellschaft},
institution = {Deutsche Arbeitsgemeinschaft für Mustererkennung e.V.},
school = {Biologische Kybernetik},
address = {Berlin, Germany},
month = sep,
year = {2004},
slug = {2644},
author = {Sinz, F. and Candela, JQ. and BakIr, G. and Rasmussen, CE. and Franz, M.},
month_numeric = {9}
}