Cone-restricted kernel subspace methods

Kobayashi, Takumi; Yoshikawa, Fumito; Otsu, Nobuyuki

doi:10.1109/icip.2010.5653014

Cited by 5 publications

(4 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This characteristic does not allow the combination of CNN features with negative coefficients; accordingly, a set of CNN features forms a convex cone instead of a subspace in a high dimensional vector space, as described in Sec.II-C. For example, it is well known that a set of front-facing images under various illumination conditions forms a convex cone, referred to as an illumination cone [16], [17]. Several previous studies have demonstrated the advantages of convex cone representation compared with subspace representation [18], [19]. These advantages naturally motivated us to replace a subspace with a convex cone in models of a set of CNN features.…”

Section: Introductionmentioning

confidence: 99%

A Method Based on Convex Cone Model for Image-Set Classification With CNN Features

Sogi

Nakayama

Fukui

2018

2018 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

In this paper, we propose a method for imageset classification based on convex cone models, focusing on the effectiveness of convolutional neural network (CNN) features as inputs. CNN features have non-negative values when using the rectified linear unit as an activation function. This naturally leads us to model a set of CNN features by a convex cone and measure the geometric similarity of convex cones for classification. To establish this framework, we sequentially define multiple angles between two convex cones by repeating the alternating least squares method and then define the geometric similarity between the cones using the obtained angles. Moreover, to enhance our method, we introduce a discriminant space, maximizing the between-class variance (gaps) and minimizes the within-class variance of the projected convex cones onto the discriminant space, similar to a Fisher discriminant analysis. Finally, classification is based on the similarity between projected convex cones. The effectiveness of the proposed method was demonstrated experimentally using a private, multi-view hand shape dataset and two public databases.

show abstract

Section: Introductionmentioning

confidence: 99%

A Method Based on Convex Cone Model for Image-Set Classification With CNN Features

Sogi

Nakayama

Fukui

2018

2018 International Joint Conference on Neural Networks (IJCNN)

View full text Add to dashboard Cite

show abstract

“…These results indicate that the positive charges on surface of the jet can efficiently drive the PBIEM segments with bromoester moiety to the fiber surface. To obtain "shortened nanofibers", water/hexane mixture was used as disperse medium to preserve the electrospun nanofibers in excellent suspension within the interface [3]. Cooling the system with an ice-water bath leads to enhance the cutting efficiency.…”

Section: Resultsmentioning

confidence: 99%

Fabrication of Shortened Electrospun Fibers with Concentrated Polymer Brush Toward Biomaterial Applications

Huang

Yoshikawa²,

Zhang

et al. 2011

AMR

View full text Add to dashboard Cite

In this research, we aimed to develop a new type of core-shell electrospun fiber, possessing “short” length (core) and “concentrated” polymer brush (shell). We prepared electrospun fibers with initiating moiety for surface-initiated atom transfer radical polymerization (SI-ATRP), one of living radical polymerizations. Then we grafted poly(sodium styrene sulfonate) (PSSNa) on the fibers by SI-ATRP. After the polymerization, we mechanically cut the electrospun fibers with a homogenizer, yielding regulated shortened fibers. Our ultimate goal is to make a novel shortened nanofibril biomaterial with concentrated brush as cell growth scaffolds. Therefore we expect that this unprecedented short nanofiber can be broadly applied as a biomaterial owing to the unique structures and properties of the concentrated brush. The details will be discussed.

show abstract

“…The illumination cone is a more strict representation than the illumination subspace mentioned above. Several previous studies have demonstrated the advantages of convex cone representation compared with subspace representation (Kobayashi and Otsu, 2008;Kobayashi et al, 2010;Wang et al, 2017Wang et al, , 2018. These advantages naturally motivated us to replace a subspace with a convex cone in models for a set of CNN features including the types of features with non-negative constraint.…”

Section: Cnnmentioning

confidence: 99%

Constrained Mutual Convex Cone Method for Image Set Based Recognition

Sogi

Zhu

Xue

et al. 2019

Preprint

View full text Add to dashboard Cite

In this paper, we propose a method for imageset classification based on convex cone models. Image set classification aims to classify a set of images, which were usually obtained from video frames or multi-view cameras, into a target object. To accurately and stably classify a set, it is essential to represent structural information of the set accurately. There are various representative image features, such as histogram based features, HLAC, and Convolutional Neural Network (CNN) features. We should note that most of them have nonnegativity and thus can be effectively represented by a convex cone. This leads us to introduce the convex cone representation to image-set classification. To establish a convex cone based framework, we mathematically define multiple angles between two convex cones, and then define the geometric similarity between the cones using the angles. Moreover, to enhance the framework, we introduce a discriminant space that maximizes the between-class variance (gaps) and minimizes the within-class variance of the projected convex cones onto the discriminant space, similar to the Fisher discriminant analysis. Finally, the classification is performed based on the similarity between projected convex cones. The effectiveness of the proposed method is demon-

show abstract

Cone-restricted kernel subspace methods

Cited by 5 publications

References 13 publications

A Method Based on Convex Cone Model for Image-Set Classification With CNN Features

A Method Based on Convex Cone Model for Image-Set Classification With CNN Features

Fabrication of Shortened Electrospun Fibers with Concentrated Polymer Brush Toward Biomaterial Applications

Constrained Mutual Convex Cone Method for Image Set Based Recognition

Contact Info

Product

Resources

About