Open AccessGender Prediction using Deep Learning Algorithms and Model on Images of an Individual
Author(s) -
G. Rajendra,
K. Sumanth,
C Anjali,
Gudavalli Pardhasai,
M. Supraja
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1998/1/012014
Subject(s) - computer science , autoencoder , artificial intelligence , classifier (uml) , sort , pattern recognition (psychology) , artificial neural network , deep learning , convolutional neural network , generalization , focus (optics) , iris recognition , feature extraction , feature (linguistics) , machine learning , biometrics , mathematics , information retrieval , mathematical analysis , linguistics , philosophy , physics , optics
The classification of gender based on the biometric is an old and traditional approach which are treated as a sub-branch of “Soft Computing”. But in this proposed work, works on the celebrity images dataset obtained from the Kaggle repository. The major focus of this to extract the important features from the images by training the neural network by using an autoencoder and decoder. The system extracts the feature maps to visualize the extracted features and makes some sort of interpretations to analyze the data. Gender classification using iris information is a relatively old technique, and most gender classification methods classify all iris texture features. These features are fed to the classifier, this data might be proper or improper, and if it is improper it leads to more generalization results in less accuracy. The proposed algorithm to create increase the size of the dataset implements the Data Augmentation Technique using basic image operations. Then it applies a convolution neural network for predicting the gender of the user. The model has proved practically that increasing the size of the dataset and reducing the features improves the classifiers learning performance.