Mining discriminative co-occurrence patterns for visual recognition

Yuan, Junsong; Yang, Ming; Wu, Ying

doi:10.1109/cvpr.2011.5995476

Cited by 63 publications

(56 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since the latter implies to test all keypoints in a one-vs-rest fashion, they test only on those frequent keypoints in the image dataset. In other works, Yuan et al 22 , 23 took advantage of the use of k-nearest neighbors algorithm to group visual words and building visual phrases of different lengths in order to get relevant information. In video data mining, visual phrases have also been used for obtaining the principal objects and characters in a video by clustering on viewpoint invariant configurations 13 .…”

Section: Related Workmentioning

confidence: 99%

Bag-of-visual-ngrams for histopathology image classification

et al. 2013

View full text Add to dashboard Cite

This paper describes an extension of the Bag-of-Visual-Words (BoVW) representation for image classification (IC) of histophatology images. This representation is one of the most used approaches in several high-level computer vision tasks. However, the BoVW representation has an important limitation: the disregarding of spatial information among visual words. This information may be useful to capture discriminative visual-patterns in specific computer vision tasks. In order to overcome this problem we propose the use of visual n-grams. N-grams based-representations are very popular in the field of natural language processing (NLP), in particular within text mining and information retrieval. We propose building a codebook of n-grams and then representing images by histograms of visual n-grams. We evaluate our proposal in the challenging task of classifying histopathology images. The novelty of our proposal lies in the fact that we use n-grams as attributes for a classification model (together with visual-words, i.e., 1-grams). This is common practice within NLP, although, to the best of our knowledge, this idea has not been explored yet within computer vision. We report experimental results in a database of histopathology images where our proposed method outperforms the traditional BoVWs formulation.

show abstract

Section: Related Workmentioning

confidence: 99%

Bag-of-visual-ngrams for histopathology image classification

et al. 2013

View full text Add to dashboard Cite

show abstract

“…[11,12,22,23,25,29,33]). These methods differ in the way they transform images into sets of items which can be mined out.…”

Section: Related Workmentioning

confidence: 99%

“…Reducing the size of the vocabulary is not a good option since vocabulary size is positively correlated with good performance. Differently, [32,33] mine patterns from high-level semantic features for which smaller representational space are needed. However, such techniques can work only if the high-level features are correctly detected, which is an open question.…”

Section: Related Workmentioning

confidence: 99%

“…Interestingly, since a few years, pattern mining has been shown to be a promising avenue for discovering these dependencies e.g. [11,12,23,33]. Pattern mining algorithms are used for extracting a subset of 'meaningful' groups of features (so-called patterns), such as groups of features which frequently appear together in the images of a given class.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Histograms of Pattern Sets for Image Classification and Object Recognition

Voravuthikunchai

Crémilleux

Jurie

2014

2014 IEEE Conference on Computer Vision and Pattern Recognition

View full text Add to dashboard Cite

This paper introduces a novel image representation capturing feature dependencies through the mining of meaningful combinations of visual features. This representation leads to a compact and discriminative encoding of images that can be used for image classification, object detection or object recognition. The method relies on (i) multiple random projections of the input space followed by local binarization of projected histograms encoded as sets of items, and (ii) the representation of images as Histograms of Pattern Sets (HoPS). The approach is validated on four publicly available datasets (Daimler Pedestrian, Oxford Flowers, KTH Texture and PASCAL VOC2007), allowing comparisons with many recent approaches. The proposed image representation reaches state-of-the-art performance on each one of these datasets.

show abstract

“…Method Accuracy(%) GIST-color [19] 69.50 RBoW [21] 78.60 ± 0.70 Classemes [29] 80.60 Object Bank [15] 80.90 SP [12] 81.40 SPMSM [11] 82.30 LCSR [26] 82.67 ± 0.51 SP-pLSA [1] 83.70 CENTRIST [33] 83.88 ± 0.76 HIK [32] 84.12 ± 0.52 VC + VQ [17] 85.40 LMLF [2] 85.60 ± 0.20 LPR [25] 85.81 Hybrid-Parts + GIST-color + SP [37] 86.30 CENTRIST + LLC + Boosting [36] 87.80 RSP [7] 88.10 LScSPM [6] 89.75 ± 0.50 ISPR(our approach)…”

Section: -Scene Datasetmentioning

confidence: 99%