Levenberg-Marquardt Variants in Chrominance-Based Skin Tissue Detection

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F62690094%3A18450%2F19%3A50015941" target="_blank" >RIV/62690094:18450/19:50015941 - isvavai.cz</a>

Výsledek na webu

<a href="http://dx.doi.org/10.1007/978-3-030-17935-9_9" target="_blank" >http://dx.doi.org/10.1007/978-3-030-17935-9_9</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-030-17935-9_9" target="_blank" >10.1007/978-3-030-17935-9_9</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Levenberg-Marquardt Variants in Chrominance-Based Skin Tissue Detection

Popis výsledku v původním jazyce

Levenberg-Marquardt method is a very useful tool for solving nonlinear curve fitting problems; while it is also a very promising alternative of weight adjustment in feed forward neural networks. Forcing the Hessian matrix to stay positive definite, the parameter λ also turns the algorithm into the well-known variations: steepest-descent and Gauss-Newton. Given the computation time, the results achieved by these methods surely differ while minimizing the sum of squares of errors and with an acceptable accuracy rate in skin tissue recognition. Therefore in this paper, we propose the implementation of these variations in network training by a set of tissue samples borrowed from SFA human skin database. The RGB images taken from the set are converted into YCbCr color space and the networks are individually trained by these methods to create weight arrays minimizing the error squares between the pixel values and the function output. Consisting of hands on computer keyboards, the images are analyzed to find skin tissues for achieving high accuracy with low computation time and for comparison of the methods.

Název v anglickém jazyce

Levenberg-Marquardt Variants in Chrominance-Based Skin Tissue Detection

Popis výsledku anglicky

Levenberg-Marquardt method is a very useful tool for solving nonlinear curve fitting problems; while it is also a very promising alternative of weight adjustment in feed forward neural networks. Forcing the Hessian matrix to stay positive definite, the parameter λ also turns the algorithm into the well-known variations: steepest-descent and Gauss-Newton. Given the computation time, the results achieved by these methods surely differ while minimizing the sum of squares of errors and with an acceptable accuracy rate in skin tissue recognition. Therefore in this paper, we propose the implementation of these variations in network training by a set of tissue samples borrowed from SFA human skin database. The RGB images taken from the set are converted into YCbCr color space and the networks are individually trained by these methods to create weight arrays minimizing the error squares between the pixel values and the function output. Consisting of hands on computer keyboards, the images are analyzed to find skin tissues for achieving high accuracy with low computation time and for comparison of the methods.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ISBN

978-3-030-17934-2

ISSN

0302-9743

e-ISSN

—

Počet stran výsledku

12

Strana od-do

87-98

Název nakladatele

Springer Verlag

Místo vydání

Berlin

Místo konání akce

Granada

Datum konání akce

8. 10. 2019

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—