DMCA
P.: Cascaded pose regression (2010)
| Venue: | In: IEEE Conference on Computer Vision and Pattern Recognition |
| Citations: | 50 - 2 self |
BibTeX
@INPROCEEDINGS{Welinder10p.:cascaded,
author = {Peter Welinder and Pietro Perona},
title = {P.: Cascaded pose regression},
booktitle = {In: IEEE Conference on Computer Vision and Pattern Recognition},
year = {2010}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
We present a fast and accurate algorithm for comput-ing the 2D pose of objects in images called cascaded pose regression (CPR). CPR progressively refines a loosely spec-ified initial guess, where each refinement is carried out by a different regressor. Each regressor performs simple image measurements that are dependent on the output of the pre-vious regressors; the entire system is automatically learned from human annotated training examples. CPR is not re-stricted to rigid transformations: ‘pose ’ is any parameter-ized variation of the object’s appearance such as the de-grees of freedom of deformable and articulated objects. We compare CPR against both standard regression techniques and human performance (computed from redundant human annotations). Experiments on three diverse datasets (mice, faces, fish) suggest CPR is fast (2-3ms per pose estimate), accurate (approaching human performance), and easy to train from small amounts of labeled data. 1.
Keyphrases
pose regression human performance pre-vious regressors pose estimate accurate algorithm standard regression technique entire system spec-ified initial guess simple image measurement small amount different regressor articulated object parameter-ized variation object appearance redundant human annotation rigid transformation training example diverse datasets
