Age and gender detection using deep learning

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 01 | Jan 2023 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 0.422 | ISO 9001:2008 Certified Journal | Page 76
Age and gender detection using deep learning
1,2,3B.Tech Scholars, Department of Computer Science and Engineering, SNIST, Hyderabad-501301, India
4Assosiate Professor, Department of Computer Science and Engineering, SNIST, Hyderabad- 501301, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - The extraction of auxiliary data from various
biometric approaches, includingfingerprints, faces, iris, palms,
voices, etc., is currently the subject of research. Gender, age,
beard, mustache, scars, height, hair, skin tone, glasses, weight,
facial scars, tattoos, and other traits are all included in this
data. Each piece of information acquires relevance during
identification. One of the most important developments in
facial recognition is the ability to determine a person's age
and gender. Given the significance of age and gender in social
interactions, it might be challenging to infer these two facial
characteristics from a single-face photo. The term "computer
vision" refers to the several terminologies used to scan images
and determine an individual's age and gender.
Key Words: Age, Gender, Detection, Features,
Extraction, Data Visualization, Classify, CNN, proposed
framework, Training data.
I. INTRODUCTION
In recent years, data from a human face has been used in
numerous real-world applications, including social
networking, security monitoring, advertising, and
entertainment. Automatic age and gender prediction from
face images is a key field for computer vision researchers
because it plays a crucial role in interpersonal
communication. Face analysis has attracted the attention of
academics in fields including demographic data gathering,
surveillance, human-computer interaction, marketing
intelligence, and security because face analysis includes a
crucial component called face age and gender recognition.
Both healthy and ill people are paying more heedtonutrition
advice lately. [5] This essay focuses on giving dietary
recommendations to people according to their age and
gender. Numerous techniques exist for identifyingaperson's
gender based on theirbiologicalcharacteristics,mannerisms,
and behaviors. The features ofapersoncangiveawayprecise
information about them, including their age, gender, mood,
ethnicity, and expression. Gender identification from a
person's facial image is a difficult application in the field of
computer vision, image analysis, and artificial intelligence,
which categorizes gender based on masculinity and
femininity. A binary classification problem leads to the
assignment of a gender category to an individual. Gender
identification is one element of facial analysis that
concentrates on classifying the images in a controlled
environment. According to, genderclassificationisnecessary
in an uncontrolled context. The gender of a person gives
additional information that makes retrieval quick and
accurate. It is crucial to identify the data present in the
photos. For the purposes of detection, thedatathattheimage
includes must be altered and manipulated. Different kinds of
strategies are used for both problem detection and solution.
In a facial recognition method: There is a wealth of
information contained in the looks on the faces. There are
numerous expressions involved wheneverapersoninteracts
with another person. Calculating certain parameters ismade
easier by modifying expressions.Ageestimateisamulti-class
problem in which different categories of years are used. It is
challenging to compile the photographs because people of
different ages have different facial features. There are
numerous age-detecting techniques. The preparation is
applied to the image. The convolution network is used to
extract features from the neural network. The image is then
assigned to one of the age classes based on the trained
models. The photos' features are taken out for additional
processing. After more processing, the features are
transmitted to the training systems. The databases offer a
study of the facial features and aid in face detection to prove
the age of the subject in the picture.
II. LITERATURE SURVEY
The study entailsa thorough documentationofindividual
differences based on age, gender, identity, and other
characteristics. In 2001, D. Kornack and P. Rakic proposed
Adult Primate Neocortex [1]. Age estimation using
convolutional network was introduced by Chenjing Yan. The
face-matching brain activation tests are carried out and
tested outside of the scanner. In terms of facial processing,
both older and younger persons showed the same results. M.
Young with proposed identical facial perspectives in both
scenarios, performance is excellent [2]. There is no single
cause for the aging of the elderly. The accounting of such
findings is the consequence of a mix of many elements. The
findings, which are based on all credentials stored in certain
environments, need to be monitored. In this research, Hang
Qi et al [3]. Made the claim that a number of methods have
emerged for the detection of faces that can also determine a
person's age. Here, an automated system that can determine
the age and assist in differentiating between a child's and an
adult's face has been suggested. The system is composed of
three components. A. Kumar and F. Shaik’s theory about age
categorization, face alignment, and face detection are the
three [4]. The standard face detection and alignment
techniques are used to build the face samples. Eran Eidinger,
Roee Enbar, and Tal Hassner’s theory helps with unfiltered
face. [6] The local face components that are visible in the
1R.Varun Reddy, 2A.Haritha Reddy, 3Abdul Kareem, 4Dr. R. Lakshmi Priya

photos are extractedusingICA.Ithasbeendemonstratedthat
this system is substantially faster and that the outcomes are
effective. Aditya K. Saxena, Shweta Sharma and Vijay K.
Chaurasiya used curvalet domain[7]. Therefore, this system
may be used as a prototype in the future. The Conditional
Probability Neural Network (CPNN), a distributed learning
technique used for age prediction from facial expressions,
was proposed by Chao Yin et al in the paper. Age estimation
using a hierarchical classifier based on global and local facial
features [8]. The goal values and the conditional feature
vectors are utilized as the input in a three-layer neural
network-based model. This may aid in its learning of actual
ages. The neural network's link between the facial imageand
the associated label distribution is how this system learns.
According to the earlier strategy, the connection should be
applied in accordance with the maximum entropy model[9] .
CPNN has demonstrated that it outperforms all previously
developed methodologies in terms of results. Results were
easily obtained using this procedure, and there was
computationally involved andtheoutcomesareveryefficient
and good in this case.
III. IMPORT LIBRARIES
Before we actually get into the model-building phase, we
need to ensure that the right libraries and frameworks have
been installed. The below libraries are required to run this
project:
Pip install OpenCV
Python numpy
pip install pafy
pip install youtube_dl
pafy : Pafy library is used to retrieve YouTube content and
metadata.
IV. METHODOLOGY
Deep Learning
A kind of artificial intelligence (AI) called deep learning
aims to replicate the human brain by learning from
experience. These representations are learned through an
instructional process. We must first train the software witha
huge number of object photos that wecategorizeintoseveral
classes in order to teach it how to detect an object. On
average, deep learning-basedalgorithms need more training
data and take longer to train than conventional machine
learning techniques.Findingdistinctivecharacteristicswhen
attempting to identify any object or character in an image
takes effort and complexity. When applying deep learning
techniques, issues can be resolved as opposed to classical
machinelearning,wherefeaturesareautomaticallyextracted
from data. An elaborate neural network with hiddenlayersis
known as deep learning.
Convolution Neural Network (CNN):
CNN, a sort of artificial neural network, is often used for
categorizing and recognizing objects in images or pictures.
Usinga CNN, deep learning can identify items in animage.An
input layer, hidden layers, and an output layer make up a
conventional neural network. The structure of the human
brain served as inspiration for CNN. Artificial neurons, also
known as nodes, in CNNs, accept inputs, process them, and
then providethe outcomeas an output, muchlikeaneuronin
the brain does when sending information between cells. The
images are used as a source of data. There may be numerous
hidden layersinCNN’s,andeachoneemploysmathematicsto
extract features from the image. The layer at the bottom that
divides features from inputs is the convolution layer.
V. ALGORITHM:
When an image is taken from a great distance and traits
that resemble haar are used, gender recognition can be
somewhat challenging. We came up with a simple but
effective solution for this issue. We used a cascading
approach. ROI (Region of Interest) serves as our face in this
study. As shown in fig-5.1 we gave the classifier the ROI
image. We attempted to identify the female face in this
document. 500 photos of women and 500 photographs of
men were used to trainourhaarcascadeclassifier.Inorderto
practice, we used frontal face photos with external elements
like a hairdo, makeup, and accessories like earrings and
glasses. This research attempts to identify an object that has
positive XML training. In our instance, we only use good and
negative images of men and women.
Fig: 5.1 Flow chart

VI. INPUT
This study's major objective is to streamline and
accelerate a whole system. The data can be entered into the
algorithm in a number of different ways to expedite the
process. To begin, the user can immediately gather data by
using the system's webcam or similarwebcamdigital device.
VII. FACE DETECTION
A piece of software called a face recognition system
may compare a human face in a frame of video or a digital
image to a database of faces. Some natural (lighting, posing
angles, facial labeling) and digital (noise, interference)
adjustments are made when a face is detected in a frame.
The challenge of recognizing a human face is caused by two
characteristics of a human face as a template:(1)Therearea
huge and almost definitely infinite number of templates, or
faces to be classified. (2) Almost all patterns resemble one
another. To resolve this issue and make the algorithm as
effective as possible, we can use a variety of audience
records. The audience set also serves as a benchmark for
gender categorization in neural networks.
VIII. FACE PROCESSING
If a face is discovered following the face detection process.
To start processing, a convolutional neural network,orCNN,
might be employed. This particular deep neural network is
used largely for image processing. During its training phase,
CNN generates a range of estimates. It'sa typeofdeepneural
network that's frequentlyemployedinimageprocessing and
NLP. CNN will conduct the actual training phase, and many
forecasts will be made. The two genders that can be
anticipated are male and female. Estimating age is a multi-
class job where the eras are broken down into groups. It's
challenging to compare the facial features of persons of
different ages because they have to get precise data. To
expedite the process, we separated the population into
several age groups. Eight kinds of age estimations are
possible: (0- 2), (4-6), (8-12), (15-20), (25-32),(38-43),(48-
53), and (60–100).
IX. OUTPUT
Launch the path of the code directory into the
command prompt. Then the path to the location of the code
has been set. Now, execute the code using the command
python filename.py. The project window appears on the
screen which begins to identify the object in front of the
webcam. If the object is identified, the algorithm classifies
the gender type along with age group as shown in below fig-
9.1 .
Fig: 9.1 Output
X. CONCLUSIONS
This project's discoveries regarding age-estimating
contributions and gender categorizationcanbeusedtosolve
challenges in real-time applications. Although age and
gender classification concerns were addressed by earlier
systems, much of this research was previously restricted to
constrained photos taken in lab settings. The visual
discrepancies that are common in real-world photos on
social media platforms and in online archives are not
sufficiently reflected by such settings. Ontheotherhand,it is
more difficult to find photographs online because there are
so many moreofthem. We examineDeep-performanceCNNs
on these tasks using Internet data and a related field's
example of facial recognition. We present our results
utilizing a lean Deep Learning architecture that avoids
overfitting due to the absence of labeled data. In particular
current network topologies, our network is "shallow",
minimizing the number of parameters and the risk of
overfitting.
XI. ACKNOWLEDGEMENT (Optional)
We would like to express our special thanks to our
mentor Dr. R. Lakshmi Priya who gave us a golden
opportunity to do this wonderful project on this topic which
also helped us in doing a lot of research and we came to
know about so many new things. We are really thankful to
them. Secondly, we would also like to thank my friends who
helped us a lot in finalizing this project within the limited
time frame.
[1] Turabzadeh, Saeed, Hongying Meng, Rafiq M. Swash,
Matus Pleva, and Jozef Juhar. "Real-time emotional state
detection from facial expression on embedded devices."
In 2017 Seventh International ConferenceonInnovative
Computing Technology (INTECH), pp. 46- 51. IEEE,
2017.
XII. REFERENCES

[2] Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E.Hinton.
"Imagenet classification with deep convolutional neural
networks." In Advances in neural information
processing systems, pp. 1097-1105. 2012.
[3] A. Kumar and F. Shaik,” Importance of Image
Processing”, 2016.
[4] Eran Eidinger, Roee Enbar, and Tal Hassner,”Age and
gender Estimation of Unfiltered Faces”, 2014 in IEEE.
[5] Aditya K. Saxena, Shweta Sharma and Vijay K.
Chaurasiya,” Neural Network based Human Age-group
Estimation in Curvelet Domain”, 2015 The Authors
Published by Elsevier.
[6] J. Chen, S. Shan, C. He, G. Zhao, M. Pietikainen, X. Chen,
and W. Gao. Wld: “A robust local image descriptor”,
Trans.Pattern Anal. Mach. Intell., 7. S. E. Choi, Y. J. Lee, S.
J. Lee, K. R. Park, and J. Kim. “Age estimation using a
hierarchical classifier based on global and local facial
features”, Pattern Recognition, 44(6):1262–1281, 2011
[7] Hang Qi and Liqing Zhang,” Age Classification System
with ICA Based Local Facial Features”, 2009 Springer-
Verlag Berlin Heidelberg.
[8] Chenjing Yan, Congyan Lang, Tao Wang, Xuetao Du, and
Chen Zhang,” Age Estimation Based on Convolutional
Neural Network”, 2014 Springer International
Publishing Switzerland.
[9] Dehghan, Afshin, Enrique G. Ortiz, Guang Shu, and Syed
Zain Masood. "Dager: Deep age, gender and emotion
recognition using convolutional neural network." arXiv
preprint arXiv:1702.04280 (2017).

Age and gender detection using deep learning

More Related Content

Similar to Age and gender detection using deep learning (20)

More from IRJET Journal (20)

Recently uploaded (20)

Age and gender detection using deep learning