Using bagging and boosting algorithms for 3D object labeling

Herouane, Omar; Moumoun, Lahcen; Gadi, Taoufiq

doi:10.1109/iacs.2016.7476070

Cited by 3 publications

(4 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The method called the weak learner iteratively, the training data used is taken from several subsets of the entire database. A single robust classifier is then constructed by combining the resulting weak learner with the resampling training set [27]. There are many boosting variations, one of which is AdaBoost.M1, specially where ℎ 𝑡 𝛽 𝑡 is the induced classifier (with ℎ 𝑡 (𝑥) ∈ ℂ) and give…”

Section: Ensemble Methodsmentioning

confidence: 99%

Model optimisation of class imbalanced learning using ensemble classifier on over-sampling data

Kurniawati

Prabowo

2022

IJ-AI

View full text Add to dashboard Cite

<span lang="EN-US">Data imbalance is one of the problems in the application of machine learning and data mining. Often this data imbalance occurs in the most essential and needed case entities. Two approaches to overcome this problem are the data level approach and the algorithm approach. This study aims to get the best model using the pap smear dataset that combined data levels with an algorithmic approach to solve data imbalanced. The laboratory data mostly have few data and imbalance. Almost in every case, the minor entities are the most important and needed. Over-sampling as a data level approach used in this study is the synthetic minority oversampling technique-nominal (SMOTE-N) and adaptive synthetic-nominal (ADASYN-N) algorithms. The algorithm approach used in this study is the ensemble classifier using AdaBoost and bagging with the classification and regression tree (CART) as learner-based. The best model obtained from the experimental results in accuracy, precision, recall, and f-measure using ADASYN-N and AdaBoost-CART.</span>

show abstract

Section: Ensemble Methodsmentioning

confidence: 99%

Model optimisation of class imbalanced learning using ensemble classifier on over-sampling data

Kurniawati

Prabowo

2022

IJ-AI

View full text Add to dashboard Cite

show abstract

“…Table 2 illustrates the recognition statistical results using Boosted-SVM, Multiclass-SVM, AdaBoost.M1 [13] classifiers under different training samples and this of Kalogerakis et al [14] using the JointBoost classifier.…”

Section: Class Name Segments/ Objectmentioning

confidence: 99%

“…Each part of a 3D object is represented by the distribution of its faces areas with respect to their shape index values. The choice of the shape index is done as it allows the shape of a 3D object surface to be defined [13].…”

Section: D Object-parts Recognition Processmentioning

confidence: 99%

“…Moreover, many researches works using boosting algorithms for recognizing components of scanned 3D objects were developed. In one of our early works [13], we conceived a framework for 3D object-parts learning in which we used Ababoost.M1 with decision-stump as a weak learner. The obtained results are very encouraging however, the system still find some difficulties concerning recognizing some heterogeneous classes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Hybrid Boosted-SVM Classifier for Recognizing Parts of 3D Objects

Herouane¹,

Moumoun²,

Gadi³

et al. 2018

IJIES

Self Cite

View full text Add to dashboard Cite

Abstract:The main objective of classifying parts of a 3D object is to group them into similar and meaningful categories. This purpose is a challenging task in pattern recognition, especially when the risk of assigning simultaneously the same label of a class to many homogenous categories is high. Hence, this paper aims to develop a robust algorithm for 3D objects-parts recognition based on a hybrid approach combining a boosting algorithm and a SVM classifier. This approach called Boosted-SVM consists of two steps, the first one is training the strong SVM classifier to be a weak base learner, and the second one is combing these trained classifiers into a strong one using the Adaboost algorithm. The training of the classification model is based on a significant and robust feature vector representing parts of the 3D object, the Shape Spectrum Descriptor (SSD) is used for this purpose. The performed experimental results have shown excellent recognition rate and very short training time of the built system.

show abstract