Learning-Based Local Visual Representation and Indexing by Rongrong Ji, Yue Gao, Ling-Yu Duan, Hongxun Yao, Qionghai

By Rongrong Ji, Yue Gao, Ling-Yu Duan, Hongxun Yao, Qionghai Dai

Learning-Based neighborhood visible illustration and Indexing, studies the state of the art in visible content material illustration and indexing, introduces state-of-the-art concepts in studying established visible illustration, and discusses rising subject matters in visible neighborhood illustration, and introduces the latest advances in content-based visible seek techniques.

  • Discusses cutting-edge tactics in learning-based neighborhood visible representation.
  • Shows find out how to grasp the fundamental options wanted for development a large-scale visible seek engine and indexing system
  • Provides perception into how desktop studying innovations should be leveraged to refine the visible acceptance method, specially within the a part of visible function representation.

Show description

Read or Download Learning-Based Local Visual Representation and Indexing PDF

Similar computer vision & pattern recognition books

Advances in Geometric Modeling and Processing: 6th International Conference, GMP 2010, Castro Urdiales, Spain, June 16-18, 2010, Proceedings

This ebook constitutes the refereed court cases of the sixth foreign convention on Geometric Modeling and Processing, GMP 2010, held in Castro Urdiales, Spain, in June 2010. The 20 revised complete papers offered have been rigorously reviewed and chosen from a complete of 30 submissions. The papers conceal a large spectrum within the sector of geometric modeling and processing and handle subject matters equivalent to ideas of transcendental equations; quantity parameterization; soft curves and surfaces; isogeometric research; implicit surfaces; and computational geometry.

Discrete Geometry for Computer Imagery: 15th IAPR International Conference, DGCI 2009, Montréal, Canada, September 30 - October 2, 2009, Proceedings

This publication constitutes the refereed complaints of the fifteenth IAPR foreign convention on Discrete Geometry for machine Imagery, DGCI 2009, held in Montr? al, Canada, in September/October 2009. The forty two revised complete papers have been conscientiously reviewed and chosen from a number of submissions. The papers are geared up in topical sections on discrete form, illustration, acceptance and research; discrete and combinatorial instruments for photo segmentation and research; discrete and combinatorial Topology; versions for discrete geometry; geometric transforms; and discrete tomography.

Independent Component Analysis of Edge Information for Face Recognition

The publication offers examine paintings on face reputation utilizing area info as good points for face acceptance with ICA algorithms. The self reliant elements are extracted from aspect info. those self sufficient parts are used with classifiers to check the facial pictures for popularity objective. of their examine, authors have explored Canny and LOG area detectors as usual aspect detection equipment.

Advanced Technologies in Ad Hoc and Sensor Networks: Proceedings of the 7th China Conference on Wireless Sensor Networks

Complicated applied sciences in advert Hoc and Sensor Networks collects chosen papers from the seventh China convention on instant Sensor Networks (CWSN2013) held in Qingdao, October 17-19, 2013. The e-book positive aspects state of the art experiences on Sensor Networks in China with the subject of “Advances in instant sensor networks of China”.

Additional resources for Learning-Based Local Visual Representation and Indexing

Sample text

Although our CASL detector produces generally fewer salient regions compared with the alterative approaches, we still achieve comparable performance by including semi-local spatial layouts in detection. Such spatial layout prevents unstable detections that are frequently found in traditional local feature detectors. 6 shows the quantitative evaluations of the detector repeatability comparisons in the sequences of different scales, viewpoints, blurs, compressions, and illuminations. We can see that, in many cases, our CASL detector produces more repeatable detection results in the repeatability comparisons of illuminations (fifth subfigure) and viewpoints (sixth subfigure); in all cases better in the repeatability comparisons of compressions (fourth subfigure); and in some cases better in the repeatability comparisons of blurs (third subfigure).

6 While the Operation Array is not empty {Get the first element nj 7 If Frej ≤ ξmin { sibling If there are sibling leaves nj of this node{ 9 Leaf Delete, push all sibling leaves (nsj ) into Operation Array} 10 Else { Parent Withdraw, push nj ’s parent as a renewed leaf into Operation Array}} 11 If Frej ≥ ξmin { 8 split Leaf Split, push new leaves {nj } into Operation Array} 13 Delete nj } 14 Output: Refined vocabulary tree after adaption. 12 the original database need not be updated.

This step naturally enables learning-based detection by integrating category learning into the mean shift weights and kernels. We conducted quantitative comparisons on image search, object categorization, and detector repeatability evaluations. We compared our performance with current approaches, based on which we further discussed the suitable and unsuitable scenarios for deploying our CASL detector. Two interesting questions remain: First, we would further integrate category learning into the DoCG construction phase, in which we can adopt category-aware distributions to supervise the contextual representation of local features.

Download PDF sample

Rated 4.55 of 5 – based on 48 votes

Related posts