Recognition with local features: the kernel recipe

By Christian Wallraven, Barbara Caputo, Arnulf B. A. Graf. Proceedings Ninth IEEE International Conference on Computer Vision, 2003

440 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/c393b31ca71e8c4dd7c8c5a11653b18447c90466

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

A Machine Learning Approach to Android Malware Detection (2012, methodology): “In [34], a simple method was proposed: The idea is to ﬁnd the best match y j ∈ y for each x i ∈ x , and then the best x i for each y i , and to sum the k 0 ( x i , y j ) for each of these matches.”
The pyramid match kernel: discriminative classification with sets of image features (2005, methodology): “In contrast, previous approaches have used each feature in a set to independently index into the second set; this ignores possibly useful information that is inherent in the cooccurrence of a set of distinctive features, and it fails to distinguish between instances where an object has varying numbe”
Object Classification from a Single Example Utilizing Class Relevance Metrics (2004, background): “Recently attempts have been made to tackle this problem [19], we choose to follow [20] by adopting the principal angles kernel approach that implicitly maps x of size n (cid:2) k to a signiﬁcantly higher -dimensional feature space (cid:30) ( x ) 2 F .”
Gaussian Processes for Object Categorization (2010, methodology): “Several authors have explored correspondence-based kernels (Zhang et al. 2006; Wallraven et al. 2003), where the distance between a set of local feature descriptors— potentially including appearance and shape/position—is computed based on associating pairs of descriptors.”
Pose pooling kernels for sub-category recognition (2012, background): “Ideally, we first consider a match kernel in the spirit of [32], which would compare two sets of poselet activations by transforming each poselet detection in one image to another poselet detection in a second image, and then comparing the corresponding image descriptors.”
Generic object recognition with boosting (2006, background): “Wallraven et al. [40] use support vector machines combined with local features for object recognition.”
Active Learning with Gaussian Processes for Object Categorization (2007, methodology): “Figure 1 shows the proposed framework for the image categorization task.”
Sparse Modeling for Image and Vision Processing (2014, methodology): “Similar schemes have also been used in image categorization [Wallraven et al., 2003].”
Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories (2006, background): “For such whole-image categorization tasks, bag-offeatures methods, which represent an image as an orderless collection of local features, have recently demonstrated impressive levels of performance [5, 21, 22, 23].”
A Discriminative Kernel-Based Approach to Rank Images from Text Queries (2008, methodology): “This parameterization can be extended to nonlinear mappings through the kernel trick, which allows PAMIR to benefit from effective picture kernels recently introduced in the computer vision literature, e.g., [ 48 ], [27], and [30].”
A Statistical Approach to Material Classification Using Image Patch Exemplars (2009, background): “with some minor modifications, can be thought of as a Mercer kernel [55].”
The Pyramid Match Kernel: Efficient Learning with Sets of Features (2007, methodology): “Other approaches to this problem have recently been proposed (Wallraven et al. 2003; Lyu 2005; Boughhorbel et al. 2004; Kondor and Jebara 2003; Wolf and Shashua 2003; Moreno et al. 2003; Shashua and Hazan 2005; Cuturi and Vert 2005; Boughorbel et al. 2005; Lafferty and Lebanon 2002), but unfortunate”
Graphical Gaussian Vector for Image Categorization (2012, methodology): “Finding an explicit correspondence between local features is another way of categorizing images using a BoW [4], [12], [26], and this approach has been improved by representing a spatial layout of local features as a graph [11], [2], [16], [8].”
Classifying materials in the real world (2010, background): “Another lively direction of research for visual kernels addresses the issue of how to use local features as input of an SVM classifier [34,33].”