Learning Local Convolutional Features for Face Recognition with 2D-Warping
Abstract
The field of face recognition has seen a large boost in performance by applying Convolutional Neural Networks (CNN) in various ways. In this paper we want to leverage these advancements for face recognition with 2D-Warping. The latter has been shown to be effective especially with respect to pose-invariant face recognition, but usually relies on hand-crafted dense local feature descriptors. In this work the hand-crafted descriptors are replaced by descriptors learned with a CNN. An evaluation on the CMU-MultiPIE database shows that in this way the classification performance can be increased by a large margin.
Cite
Text
Hanselmann and Ney. "Learning Local Convolutional Features for Face Recognition with 2D-Warping." European Conference on Computer Vision, 2016. doi:10.1007/978-3-319-49409-8_62Markdown
[Hanselmann and Ney. "Learning Local Convolutional Features for Face Recognition with 2D-Warping." European Conference on Computer Vision, 2016.](https://mlanthology.org/eccv/2016/hanselmann2016eccv-learning/) doi:10.1007/978-3-319-49409-8_62BibTeX
@inproceedings{hanselmann2016eccv-learning,
title = {{Learning Local Convolutional Features for Face Recognition with 2D-Warping}},
author = {Hanselmann, Harald and Ney, Hermann},
booktitle = {European Conference on Computer Vision},
year = {2016},
pages = {747-758},
doi = {10.1007/978-3-319-49409-8_62},
url = {https://mlanthology.org/eccv/2016/hanselmann2016eccv-learning/}
}