Towards optimal bag-of-features for object categorization and semantic video retrieval

By Yu-Gang Jiang, Chong-Wah Ngo, Jun Yang. 2007

724 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Information Retrieval & Search

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 Information Retrieval & Search

Citations in the corpus

This looks like that: deep learning for interpretable image recognition (2018, background): “It relates to traditional “bag-of-visual-words” models used in image recognition [21, 6, 17, 38, 31].”
Unified Video Annotation via Multigraph Learning (2009, background): “1There are also several other such public baselines, such as VIREO-374 [13] and Mediamill-101 [29].”
An Efficient Dense and Scale-Invariant Spatio-Temporal Interest Point Detector (2008, methodology): “Then, a bag-of-words is computed for each video using a weighted approach, similar to the one proposed in [19], where each visual word t k of the visual vocabulary T = [ t 1 , . . . , t k , . . . , t K ] is given the weight with M i the number of interest points that have visual word k as their i…”
Visual Word Ambiguity (2010, background): “Tuytelaars and Schmid [28] and Jiang et al. [11] assign an image feature to visual words that are neighbors in feature space.”
Asymmetric Minwise Hashing for Indexing Binary Inner Products and Set Containment (2015, background): “Thus in practice, often only the presence or absence information suffices [9, 20, 24].”
Evaluating Color Descriptors for Object and Scene Recognition (2010, background): “In general, systems for category recognition on images [1], [2], [3], [4], [5] and video [6], [7], [ 8 ] use machine learning based on image descriptions to distinguish object and scene categories.”
Consumer video understanding: a benchmark database and an evaluation of human and machine performance (2011, methodology): “For all the three features, a softweighting scheme is employed to alleviate the quantization effects in generating bag-of-X features [9].”
Time-series classification using neural Bag-of-Features (2017, background): “However, it was established that it can be used for a wide range of tasks, such as video [11], and time-series representation [7]–[9].”
Feature sampling and partitioning for visual vocabulary generation on large action classification datasets (2014, methodology): “Despite extensive work showing various aspects in which BoF may be improved to squeeze out additional performance [5,6,7,4,8,9,10], there remains interesting questions yet to be evaluated empirically, particularly when considering huge action classification datasets with a large number of classes [1”
Evaluating bag-of-visual-words representations in scene classification (2007, methodology): “There have been works using this vectorquantized keypoint feature, or bag-of-visual-word representation, for image classification [6, 7, 3, 14, 20, 19].”
The Visual Extent of an Object (2011, methodology): “The χ 2 kernel was found to be the best choice for the kernel function (Jiang et al. 2007; Zhang et al. 2007).”
Beyond Pairwise: Provably Fast Algorithms for Approximate k-Way Similarity Search (2013, background): “It is often the case, in BoW representations, that just the presence or absence (0/1) of specific feature words captures sufficient information [7, 16, 20], especially with (e.”
High-level event recognition in unconstrained videos (2012, methodology): “[55] conducted a series of analysis on several choices of”
Semi-Supervised Hashing for Large-Scale Search (2012, methodology): “Particularly, a visual vocabulary with 500-length code book and a soft assignment strategy was used for deriving the image features, as described in [36].”
UvA-DARE ( Digital Academic Repository ) The visual extent of an object : suppose we know the object locations (2011, methodology): “The χ2 kernel was found to be the best choice for the kernel function (Jiang et al. 2007; Zhang et al. 2007).”
Concept-Based Video Retrieval (2009, methodology): “The classical method to obtain separate classifiers is to learn the same concept detector on several unimodal [109, 240, 257, 267, 286] or multimodal features [135, 228, 272].”