Multiple kernels for object detection

By Andrea Vedaldi, Varun Gulshan, Manik Varma, Andrew Zisserman. 2009

868 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Object Detection via a Multi-region and Semantic Segmentation-Aware CNN Model (2015, neutral): “This simple pipeline surpasses by a large margin the detection performance of all the previously published systems, such as deformable parts models [8] and non-linear multi-kernel approaches [28].”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, background): “This hypothesis is grounded, for example, in the wide range of papers that attempt to boost detection accuracy with work along four axes: (1) rich structured models [20, 42]; (2) multiple feature learning [38, 41]; (3) learned histogram-based features [11, 29, 32]; or (4) unsupervised feature learni”
Detect What You Can: Detecting and Representing Objects Using Holistic Models and Body Parts (2014, background): “This is evident from the low performance of state-of-the-art object detectors [9, 26, 24] for animals, which are highly deformable objects.”
On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation (2015, methodology): “Example case: χ(2) kernel The χ(2) kernel function has been applied with good performance in various image classification tasks [41, 54, 55].”
Learning from Distributions via Support Measure Machines (2012, methodology): “For the SVM, we adopt a Gaussian RBF kernel with χ2-distance between the histograms (Vedaldi et al. 2009), i.e., K(hi,hj) = exp ( −γχ2(hi,hj) ) where χ2(hi,hj) = ∑M r=1 (hi(cr)−hj(cr)) 2 hi(cr)+hj(cr) .”
Ensemble of exemplar-SVMs for object detection and beyond (2011, methodology): “It is worthwhile pointing out some of the key differences between our approach and other related SVM-based techniques such as one-class SVMs [18, 5], multi-class kernel SVMs, kernel-learning approaches [20], and the KNNSVM algorithm [22].”
Bag of visual words and fusion methods for action recognition: Comprehensive study and good practice (2014, neutral): “Keywords Action recognition · Bag of Visual Words · Fusion methods · Survey”
R-CNN minus R (2015, methodology): “The two-stage computation using region proposal generation is a form of cascade detector [20] or jumping window [17, 19].”
DeepID-Net: multi-stage and deformable deep convolutional neural networks for object detection (2014, methodology): “The features extracted by the AlexNet are then used for learning 200 one-versus-all SVM classifiers for 200 classes.”
Modeling Spatial-Temporal Clues in a Hybrid Deep Learning Framework for Video Classification (2015, neutral): “SVM-MKL is only slightly better than the simple early and late fusion methods, which is consistent with observations in recent works on visual recognition [42].”
Fast, Accurate Detection of 100,000 Object Classes on a Single Machine (2013, background): “Traditionally, object detection is reduced to binary classification and a sliding window classifier is applied at all positions, scales and orientations of an image [20, 4, 8, 19, 22].”
Generating Object Segmentation Proposals Using Global and Local Search (2014, background): “A dominant paradigm in object detection and recognition has been to use so called bag-of-visual-words features or histograms of oriented gradients in a sliding window framework [17] [7].”
Fisher and VLAD with FLAIR (2014, background): “Sub-window search [15] and selective search [24] opened the door to the use of locality with BoW, which is computationally more expensive than HOG [8] but superior in the quality of the semantic interpretation [15, 27, 12, 24, 25, 5].”
Berkeley MHAD: A comprehensive Multimodal Human Action Database (2013, neutral): “[29] Figure 4.”
Large-scale image categorization with explicit data embedding (2010, background): “A possible criterion to find an optimal ϕ is to minimize: ∑M i=1 ∑M j=1 (K(xi, xj) − ϕ(xi)′ϕ(xj)) 2 (2) under ∑M i=1 ϕe(xi)ϕf (xi) = 0 for e 6= f (3) where ϕe(xi) denotes the e-th dimension of ϕ(xi).”
Constrained parametric min-cuts for automatic object segmentation (2010, background): “Object detectors usually sc an a large number of bounding boxes in sliding window schemes [8], [9] without considering the plausibility of pixel grou ping within each.”
What Does Classifying More Than 10, 000 Image Categories Tell Us? (2010, methodology): “In many cases such localization of objects might be useful for improving classification, but even the most efficient of the state of the art techniques [7,20,31] take orders of magnitude more time per image than the ones we consider in this study, and thus cannot be utilized given the scale of our e”
Bottom-Up Segmentation for Top-Down Detection (2013, neutral): “2 UOCTTI LSVM MDPM [8] 52.”
Object detection and segmentation from joint embedding of parts and pixels (2011, methodology): “We present a new framework in which image segmentation, figure/ground organization, and object detection all appear as the result of solving a single grouping problem.”
Selective Search for Object Recognition (2013, positive): “However, we use a more powerful (and expensive) implementation than [16, 17, 34] by employing a variety of colour-SIFT descriptors [32] and a finer spatial pyramid division [18].”