Female facial beauty analysis for assesment of facial attractivness

This paper presents a hybrid approach to estimate female facial beauty based on Machine Learning techniques. We use a combination of two approaches: Beauty Mask and Facial Proportions, to find the features that constitute Ideal Female facial beauty and thus, develop a female facial beauty scoring sy...

Full description

Saved in:
Bibliographic Details
Published in2013 2nd International Conference on Information Management in the Knowledge Economy pp. 156 - 160
Main Authors Rizvi, Qaim Mehdi, Karawia, A. A., Kumar, Sumit
Format Conference Proceeding
LanguageEnglish
Published Chitkara University 01.12.2013
Subjects
Auto-beautification;Beauty Mask
Beauty Score
Conferences
Correlation
Correlation coefficient
Face
Facial Proportions
Mouth
Vectors
Online AccessGet full text

Cover

More Information
Summary:This paper presents a hybrid approach to estimate female facial beauty based on Machine Learning techniques. We use a combination of two approaches: Beauty Mask and Facial Proportions, to find the features that constitute Ideal Female facial beauty and thus, develop a female facial beauty scoring system based on the same. The dataset used in this work consists of 30 images being rated by 29 people. These are the front facial images of Winners, 1 st Runner-up and 2 nd Runner-up of Miss Universe Beauty Pageant from 2002 to 2011. Images are represented by a 50 element vector consisting of control points being selected manually with reference to the Beauty Mask. These points are used to calculatea total of 12 distances and 7 ratios for each image. These distances and ratios are also calculated for the Beauty Mask, and the final score is given on the basis of similarity between the respective ratios. A correlation of 67.78% shows the validity of our approach. Using this approach, an application is programmed to give scores to input facial images. Apart from scoring, the application provides two separate features: First, some suggestions to improve the facial beauty for the input image and second, an auto-beautified image of the face. The distinguished image dataset, high correlation and additional features make the approach worth.