Learning Local Feature Descriptors Using Convex Optimisation
Learning Local Feature Descriptors Using Convex Optimisation
复制标题
DOI：
10.1109/tpami.2014.2301163
复制
发表时间：
2014-08-01
影响因子：
23.6
通讯作者：
Zisserman, Andrew
中科院分区：
文献类型：
--
作者：
Simonyan, Karen;Vedaldi, Andrea;Zisserman, Andrew
这项工作的目的是学习描述符适用于稀疏特征检测器中使用的视点不变匹配。我们为实现这一目标做出了许多新颖的贡献。首先，研究表明，学习描述符的池化区域可以用公式表示为使用稀疏性选择区域的凸优化问题。其次，它表明，描述符降维也可以制定为一个凸优化问题，使用Mahalanobis矩阵核范数正则化。这两种公式都是基于区分性大幅度学习约束。作为第三个贡献，我们评估的压缩描述符的性能，从学习的实值描述符二进制化。最后，我们提出了一个扩展我们的学习配方弱监督的情况下，这使我们能够学习的描述符从未注释的图像集合。事实证明，新的学习方法在Brown等人[3]的注释局部补丁数据集和Philbin等人[22]的未注释照片集的描述符学习方面优于现有技术。
The objective of this work is to learn descriptors suitable for the sparse feature detectors used in viewpoint invariant matching. We make a number of novel contributions towards this goal. First, it is shown that learning the pooling regions for the descriptor can be formulated as a convex optimisation problem selecting the regions using sparsity. Second, it is shown that descriptor dimensionality reduction can also be formulated as a convex optimisation problem, using Mahalanobis matrix nuclear norm regularisation. Both formulations are based on discriminative large margin learning constraints. As the third contribution, we evaluate the performance of the compressed descriptors, obtained from the learnt real-valued descriptors by binarisation. Finally, we propose an extension of our learning formulations to a weakly supervised case, which allows us to learn the descriptors from unannotated image collections. It is demonstrated that the new learning methods improve over the state of the art in descriptor learning on the annotated local patches data set of Brown et al. [3] and unannotated photo collections of Philbin et al. [22].