Object categorization by learned universal visual dictionary

By John M. Winn, Antonio Criminisi, Tom Minka. 2005

979 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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

In Media Technology

Citations in the corpus

Aggregating Local Image Descriptors into Compact Codes (2012, methodology): “in [30] for GMM modeling and in [31], [32] for better separability of linear classifiers.”
LabelMe: A Database and Web-Based Tool for Image Annotation (2008, neutral): “We compare the LabelMe dataset against four annotated datasets currently used for object detection and recognition: Caltech-101 [12], MSRC [45], CBCL-Streetscenes [5], and PASCAL2006 [9].”
Using Multiple Segmentations to Discover Objects and their Extent in Image Collections (2006, methodology): “Caltech [9] 4,090 4 + background MSRC [27] 4,325 23 object and scene categories LabelMe [18] 1,554 cars, buildings, trees”
Land-Use Scene Classification in High-Resolution Remote Sensing Images Using Improved Correlatons (2015, methodology): “Recently, the Bag-of-Visual-Words (BoVW) model initially proposed for text categorization has been successfully applied to image classification [4]–[6], natural scene categorization [7]–[9], and land-use scene classification using HRIs [10], [11].”
Shape and Appearance Context Modeling (2007, methodology): “We also tested the linear filter bank (LF) used in [26], and a combination of the LF and the HOG of the L channel of the Lab color space.”
Local Learning to Improve Bag of Visual Words Model for Facial Expression Recognition (2013, background): “Although most of the state of the art approaches are based on sparse descriptors, others have used dense descriptors (Fei-Fei & Perona, 2005; Winn et al., 2005).”
SEMANTIC SEGMENTATION OF AERIAL IMAGES IN URBAN AREAS WITH CLASS-SPECIFIC HIGHER-ORDER CLIQUES (2015, methodology): “As baseline (Winn) we train and predict multi-class labels using a multi-label Random Forest with 20 trees based on pixel-wise responses to the filterbank of (Winn et al., 2005).”
Multi-Class Segmentation with Relative Location Prior (2008, background): “There are many recent works on multi-class image segmentation that employ some kind of contextual information (e.g., He et al. 2004, 2006; Schroff et al. 2006; Shotton et al. 2006; Winn et al. 2005; Yang et al. 2007; Rabinovich et al. 2007; Shental et al. 2003; Kumar and Hebert 2005; Carbonetto et a”
A review on vision techniques applied to Human Behaviour Analysis for Ambient-Assisted Living (2012, neutral): “There are two main drawbacks to this family of techniques: 1) when using 2D silhouettes as key poses, a constraint on the point of view is added, which limits recognition to that exact point of view; to overcome this, different viewpoints can be added into the database, which can be cumbersome and l”
An Analysis of Single-Layer Networks in Unsupervised Feature Learning (2011, methodology): “K-means has been used less widely in " deep learning " work but has enjoyed wide adoption in computer vision for building codebooks of " visual words " [5] [6] [15] [31], which are used to define higherlevel image features.”
Recent advances and trends in visual tracking: A review (2011, background): “Texture is a measure of the intensity variation of a surface which quantiﬁes properties such as smoothness and regularity [34–36].”
DenseCut: Densely Connected CRFs for Realtime GrabCut (2015, methodology): “…computer vision and computer graphics applications, including image editing [LHE ∗ 07,CZM ∗ 10], object detection [SBC05], image classiﬁcation [WCM05], photo composition [CCT ∗ 09, CTM ∗ 13], scene understanding [KT ∗ 09], automatic object class discovery [ZWW ∗ 12], and ﬁne-grained…”
Random Walks for Image Segmentation (2006, methodology): “But existing literature suggests that when derived from good context-based features [18], such appearance model based differences provide a meaningful global bias for Cosegmentation [2].”
Trajectons: Action recognition through the motion analysis of tracked features (2009, methodology): “Our method proceeds according to the standard bag-ofwords approach: first, features are tracked over the video using a KLT tracker [2] to produce feature trajectories (x and y positions over time) for a number of features.”
Three things everyone should know to improve object retrieval (2012, result): “Table 3 shows a comparison of this combined method with previous results.”
Author manuscript, published in "International Journal of Computer Vision (2013)" International Journal of Computer Vision manuscript No. (will be inserted by the editor) Image Classification with the Fisher Vector: Theory and Practice (2013, positive): “We ﬁnally note that the use of the square-root transform is not speciﬁc to the FV and is also beneﬁcial to the BoV as shown for instance by Perronnin et al (2010b), Vedaldi and Zisserman (2010), Winn et al (2005).”