Efficient Mining of Frequent and Distinctive Feature Configurations

By Till Quack, Vittorio Ferrari, Bastian Leibe, Luc Van Gool. 2007

196 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Systems, Signal Processing

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

1. Luc Van Gool
2. Trevor Darrell
3. Cordelia Schmid
4. Andrew Zisserman
5. Li Fei-Fei
6. Kristen Grauman
7. Jitendra Malik
8. Xiaogang Wang

In Information Systems

1. Kristen Grauman
2. Yong Jae Lee
3. Li Fei-Fei
4. Martial Hebert
5. Élisa Fromont
6. Basura Fernando
7. Christos Faloutsos
8. Tinne Tuytelaars

In Signal Processing

1. Frédéric Jurie
2. Trevor Darrell
3. Yong Jae Lee
4. Jakob Verbeek
5. Kristen Grauman
6. Lorenzo Torresani
7. Martial Hebert
8. Élisa Fromont

Citations in the corpus

• Mining Mid-level Features for Image Classification (2014, methodology): “Applying standard relevant pattern discovery methods under these circumstances, as done in [20,34,50,53], may not be the best strategy.”
• Bundling features for large scale partial-duplicate web image search (2009, background): “[5, 14, 16, 17]) require data-dependent training to select co-occurrence features, or combinatorial feature pairing/grouping.”
• Dense interest points (2010, background): “in a freqent itemset mining context as in [21]).”
• Learning query-dependent prefilters for scalable image retrieval (2009, methodology): “An alternative is to look for discriminative metawords [19], however this has so far been applied only for class recognition of predefined classes, not to general “same/different” classification.”
• Weakly-supervised Discovery of Visual Pattern Configurations (2014, background): “Modeling the spatial and geometric relationship between co-occurring visual patterns (objects or object-parts) often improves visual recognition performance [30, 9, 25, 21, 31, 16, 8, 6, 22, 17].”
• Weakly Supervised Learning with Multi-Stream CNN-LSTM-HMMs to Discover Sequential Parallelism in Sign Language Videos (2020, unknown): “[41].”
• Shape discovery from unlabeled image collections (2009, methodology): “The BBHR measures the percentage of images in the dataset that have at least h foreground features selected, as a function of the selection threshold applied to the feature weights.”
• Multi-view object class detection with a 3D geometric model (2010, methodology): “A single spatial pyramid detector is used for the full object regions of interest (2D.3 top), while for each part region under each viewpoint, several smaller, overlapping spatial pyramid detectors are trained (2D.3 bottom).”
• Grouplet: A structured image representation for recognizing human and object interactions (2010, methodology): “Our algorithm is inspired by previous works [22, 29, 28] that also use data mining methods for feature selection.”
• Modeling spatial layout with fisher vectors for image categorization (2011, background): “Distinctive spatial configurations of visual words can also be discovered by data mining techniques, such as frequent itemsets [22].”
• Learning a hierarchy of discriminative space-time neighborhood features for human action recognition (2010, background): “The strategy of generating compound neighborhoodbased features—explored initially for static images and object recognition [34, 26, 16, 17, 25]—has since been extended to video.”
• Foreground Focus: Unsupervised Learning from Partially Matching Images (2009, methodology): “The method of Quack et al. (2007) is much more selective, as can be seen by the lower maximum BBHR that is obtained on all datasets compared with our method.”
• Unsupervised modeling of object categories using link analysis techniques (2008, methodology): “’s work [19] applies well-known data mining techniques, termed frequent itemsets and association rules, to the feature selection.”