|
Title:
|
GENDER CLASSIFICATION USING SINGLE IMAGE PER PERSON |
|
Author(s):
|
Rubata Riasat, Abdul Hannan Sadiq and Faizan Ahmad |
|
ISBN:
|
978-989-8704-41-2 |
|
Editors:
|
Katherine Blashki |
|
Year:
|
2022 |
|
Edition:
|
Single |
|
Keywords:
|
Gender Classification, Cross Gender Effect, Gender Categorization |
|
Type:
|
Full Paper |
|
First Page:
|
99 |
|
Last Page:
|
106 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
This paper presents a framework based on object appearance and shape based features to address gender classification problem through face image. Histogram of Oriented Gradients (HOG) based scheme is presented for gender classification having one frame/object for training. In presented scheme nearest neighbor based classifier Euclidean distance method is used for classification. Presented scheme's results have been compared in cross dataset environment with state-of-the-art various algebraic (PCA, LDA), geometric (Gabor Wavelets) and texture (LBP) based features. Experiments have been performed in two folds using FEI face dataset consisting of 99 objects each (Male and Female) having one frame/object. Experiment's results based on fold-2 are promising, indicating that HOG and the proposed ensemble framework have sufficient discriminative power for the gender classification problem with the accuracy of 91.41%. Current research findings include a claim that forehead, hair style, chin, ears and outer boundary of face having information about face shapes has significant importance in gender classification task. |
|
|
|
|
|
|
