Boosting k-NN for Categorization of Natural Scenes

Nock, Richard; Piro, Paolo; Nielsen, Frank; Ali, Wafa Bel Haj; Barlaud, Michel

doi:10.1007/s11263-012-0539-2

Cited by 12 publications

(22 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The boosting method has been employed in a number of computer vision applications: classification of handwritten data [3], real time face detection [44], object tracking [33], classification of trees and vehicles in urban scenes [45], categorization of natural scenes [31], keypoint detection and landmarking on human faces [10], and person re-identification [30]. It has not yet been used to evaluate point matches between overlapping 3D partial shapes, perhaps for several reasons.…”

Section: Boostingmentioning

confidence: 99%

Accurately estimating rigid transformations in registration using a boosting-inspired mechanism

Liu

Martin

et al. 2016

Pattern Recognition

View full text Add to dashboard Cite

Feature extraction and matching provide the basis of many methods for object registration, modeling, retrieval, and recognition. However, this approach typically introduces false matches, due to lack of features, noise, occlusion, and cluttered backgrounds. In registration, these false matches lead to inaccurate estimation of the underlying transformation that brings the overlapping shapes into best possible alignment. In this paper, we propose a novel boosting-inspired method to tackle this challenging task. It includes three key steps: (i) underlying transformation estimation in * Corresponding author. Email: yitian.zhao@bit.edu.cn 1 the weighted least squares sense, (ii) boosting parameter estimation and regularization via Tsallis entropy, and (iii) weight re-estimation and regularization via Shannon entropy and update with a maximum fusion rule. The process is iterated. The final optimal underlying transformation is estimated as a weighted average of the transformations estimated from the latest iterations, with weights given by the boosting parameters. A comparative study based on real shape data shows that the proposed method outperforms four other state-of-the-art methods for evaluating the established point matches, enabling more accurate and stable estimation of the underlying transformation.

show abstract

Section: Boostingmentioning

confidence: 99%

Accurately estimating rigid transformations in registration using a boosting-inspired mechanism

Liu

Martin

et al. 2016

Pattern Recognition

View full text Add to dashboard Cite

show abstract

“…The most common high-level scheme consists in gradually combining from scratch classifiers obtained at each iteration, with the objective to minimize throughout iterations a convex differentiable risk called a surrogate risk, sometimes amended with a structural part based on data [2]. Unlike so-called greedy algorithms, that repeatedly perform fine-grained optimization steps [2], boosting algorithms rely on weak optimization stages much less demanding from the statistical and computational standpoints [3], [4], [5], [6], [7]. In fact, the boosting theory involves at each iteration weak classifiers slightly different from pure random, but requires that the final combination be probably as close as required from the optimum, within polynomial time.…”

Section: Introductionmentioning

confidence: 99%

“…Nearest neighbors (NN) rules are a non-trivial field of choice for boosting algorithms [4], [5], as examples ideally play weak classifiers. In this case, we treat the boosting problem in its simplest form: the accurate leveraging of examples that vote among nearest neighbors.…”

Section: Introductionmentioning

confidence: 99%

“…No boosting results are known for these algorithms, and it is in fact not known whether they achieve convergence to the optimum of Adaboost's exponential risk. A previous approach in our line of works is algorithm UNN (for "Universal Nearest Neighbors"), which brings boosting guarantees for merely all strictly convex differentiable surrogates relevant to classification [9], [5], [6]. For a wide subset of surrogates, it yields simple and efficient estimators of posteriors [10].…”

Section: Introductionmentioning

confidence: 99%

“…Boosting compliant approximations are possible, but in the context of NN rules, they are computationally far too expensive to be performed at each boosting iteration on large datasets. Computationally affordable coarse-grained approximations are also possible, that yield compelling experimental results, but it is not known if they always lie within the boosting regime [5].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Gentle Nearest Neighbors Boosting over Proper Scoring Rules

Nock

Ali

D'Ambrosio

et al. 2015

IEEE Trans. Pattern Anal. Mach. Intell.

Self Cite

View full text Add to dashboard Cite

Abstract-Tailoring nearest neighbors algorithms to boosting is an important problem. Recent papers study an approach, UNN, which provably minimizes particular convex surrogates under weak assumptions. However, numerical issues make it necessary to experimentally tweak parts of the UNN algorithm, at the possible expense of the algorithm's convergence and performance. In this paper, we propose a lightweight alternative algorithm optimizing proper scoring rules from a very broad set, and establish formal convergence rates under the boosting framework that surprisingly compete with those known for UNN. It is an adaptive Newton-Raphson algorithm, which belongs to the same lineage as the popular Gentle Adaboost. To the best of our knowledge, no such boosting-compliant convergence rates were previously known for these algorithms. We provide experiments on a dozen domains, including the challenging Caltech and SUN computer vision databases. They display that GNNB significantly outperforms UNN, both in terms of convergence rate and quality of the solution obtained, and GNNB provides a simple and efficient contender to techniques that can be used on very large domains, like stochastic gradient descent -for which little is known to date. Experiments include a divide-and-conquer improvement of GNNB which exploits the link with proper scoring rules optimization.

show abstract

Real-Time Approaches to Computational Economics: Self Adaptive Economic Systems

Marwala

2013

Advanced Information and Knowledge Processing

View full text Add to dashboard Cite

Boosting k-NN for Categorization of Natural Scenes

Cited by 12 publications

References 36 publications

Accurately estimating rigid transformations in registration using a boosting-inspired mechanism

Accurately estimating rigid transformations in registration using a boosting-inspired mechanism

Gentle Nearest Neighbors Boosting over Proper Scoring Rules

Real-Time Approaches to Computational Economics: Self Adaptive Economic Systems

Contact Info

Product

Resources

About