K Grauman, T Darrell - Journal of Machine Learning Research, 2007 - jmlr.org
… In this work we present the pyramid matchkernel—a new kernel function over unordered feature sets that … We call the proposed matchingkernel the pyramid matchkernel because input …
K Grauman, T Darrell - … on Computer Vision (ICCV'05) Volume …, 2005 - ieeexplore.ieee.org
… in the space of vector sets. In this work we propose a pyramid matchkernel – a new kernel function over unordered featuresets that allows them to be used effectively and efficiently in …
H Sun, X Zhen, Y Zheng, G Yang… - Proceedings of the …, 2017 - openaccess.thecvf.com
… matchkernels into a global matchkernel between image sets, which is learned in a supervised way by kernel … A widely used matchkernel between two sets of feature vectors is the sum …
J Yang, J Liang, H Shen, K Wang… - IEEE transactions on …, 2018 - ieeexplore.ieee.org
… a dynamic matchkernel by adaptively calculating the matching thresholds between query … among deep CNN features. In contrast to the typical static matchkernel which is independent …
S Poullot, S Tsukatani, A Phuong Nguyen… - Proceedings of the 23rd …, 2015 - dl.acm.org
… • We adapt the kernel descriptor framework of Bo [3] to … temporal matchkernels by gathering the formalism of explicit feature … to adapt the properties of the matchingkernel and limits the …
… This section first describes the class of matchkernels that we will … a set of images indexed with ASMK\(^\star \), we cross-match … In practice, we cross match only features assigned to the …
… each image by a set of vectors (a bag-of-features), each feature representing a sub-part (a … two images is then measured by matching the similarity between the descriptor sets [1], [2]. In …
L Bo, X Ren, D Fox - Advances in neural information …, 2010 - proceedings.neurips.cc
… We observe that SIFT features in conjunction with efficient matchkernels work well on this dataset and obtain 70.8% accuracy using a single patch size, which beat SPM with the same …
… Since the matching process is asymmetric, our kernel is the average of the two matching … nearby features in one image to match nearby features in the second one while discouraging …