A structural/statistical feature based vector for handwritten character recognition

Heutte, Laurent; Paquet, Thierry; Moreau, J.V.; Lecourtier, Yves; Olivier, Christian

doi:10.1016/s0167-8655(98)00039-7

Cited by 135 publications

(68 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Three additionnal datasets on different handwritten digit recognition problems have been used: (i) the well-known MNIST database [10] with a 85 multiresolution density feature set (1 + 2 × 2 + 4 × 4 + 8 × 8) built from greyscale mean values as explained in [3]; (ii) Digits and DigReject both described in [12], on which a 330-feature set has been extracted, made from three state-of-the-art kinds of descriptors, i.e. a 117-statistical/structural feature set [13], a 128-feature set extracted from the chaincode (contour-based) [14], and the same 85-feature set as for MNIST. …”

Section: Datasetsmentioning

confidence: 99%

Influence of Hyperparameters on Random Forest Accuracy

Bernard

Heutte

Adam

2009

Lecture Notes in Computer Science

125

View full text Add to dashboard Cite

Abstract. In this paper we present our work on the Random Forest (RF) family of classification methods. Our goal is to go one step further in the understanding of RF mechanisms by studying the parametrization of the reference algorithm Forest-RI. In this algorithm, a randomization principle is used during the tree induction process, that randomly selects K features at each node, among which the best split is chosen. The strength of randomization in the tree induction is thus led by the hyperparameter K which plays an important role for building accurate RF classifiers. We have decided to focus our experimental study on this hyperparameter and on its influence on classification accuracy. For that purpose, we have evaluated the Forest-RI algorithm on several machine learning problems and with different settings of K in order to understand the way it acts on RF performance. We show that default values of K traditionally used in the literature are globally near-optimal, except for some cases for which they are all significatively sub-optimal. Thus additional experiments have been led on those datasets, that highlight the crucial role played by feature relevancy in finding the optimal setting of K.

show abstract

Section: Datasetsmentioning

confidence: 99%

Influence of Hyperparameters on Random Forest Accuracy

Bernard

Heutte

Adam

2009

Lecture Notes in Computer Science

125

View full text Add to dashboard Cite

show abstract

“…For comparison purposes, some character classifiers recognize some 97% for digits (Lee, 1996;Shouno et. al., 1999), 97% for upper-case letters and 80% for lower-case letters (Heutte, 1998). However, these results are obtained by using complex image processing techniques ( (Lee, 1996), or combination feature types, e.g.…”

Section: Resultsmentioning

confidence: 99%

“…However, these results are obtained by using complex image processing techniques ( (Lee, 1996), or combination feature types, e.g. a combination of structural and statistical features (Heutte, 1998) or complex classifiers (Shouno et. al., 1999).…”

Section: Resultsmentioning

confidence: 99%

Hybrid of HMM and Fuzzy Logic for Isolated Handwritten Character Recognition

Suliman¹

2010

Character Recognition

View full text Add to dashboard Cite

“…However, Heutte [39] gives recognition rates for well-separated uppercase characters and digits of 97.27% and 97.84%, respectively, that are improved by up to 98.54% and 98.96% if the second choice is considered as well. Mohamed and Gader present a segmentation-free system whose accuracy -with the use of lexicon -varies from 89.3% to 97.2% depending on the system.…”

Section: Discussionmentioning

confidence: 99%

An unconstrained handwriting recognition system

Kavallieratou¹,

Fakotakis²,

Kokkinakis

2002

International Journal on Document Analysis and Recognition

View full text Add to dashboard Cite

Abstract. In this paper, an integrated offline recognition system for unconstrained handwriting is presented. The proposed system consists of seven main modules: skew angle estimation and correction, printedhandwritten text discrimination, line segmentation, slant removing, word segmentation, and character segmentation and recognition, stemming from the implementation of already existing algorithms as well as novel algorithms. This system has been tested on the NIST, IAM-DB, and GRUHD databases and has achieved accuracy that varies from 65.6% to 100% depending on the database and the experiment.

show abstract

A structural/statistical feature based vector for handwritten character recognition

Cited by 135 publications

References 16 publications

Influence of Hyperparameters on Random Forest Accuracy

Influence of Hyperparameters on Random Forest Accuracy

Hybrid of HMM and Fuzzy Logic for Isolated Handwritten Character Recognition

An unconstrained handwriting recognition system

Contact Info

Product

Resources

About