Age Prediction using Image Processing

Published on 19 May 2022. Written by Aniruddha Gayake

In the film industry due to the usage of many cosmetics and modern beauty products, an older person seems like a young one. By looking at their faces we cannot easily predict their original age. But don’t worry, Python has a solution for it.

In this project, we are going to develop a system that can detect the face of an actor and predict the age. 

Read more..

SLNOTE
Project Description

To predict the age, we are going to use a convolutional neural network (CNN) architecture. This CNN uses 3 convolutional layers and 2 fully connected layers with one final output layer.

This problem can be considered as a classification problem instead of regression. The reason being estimating the exact age using regression is a challenging task. Even human beings cannot predict age just by looking at the face. So, we will try to predict the age in an age group like in 20 – 30 or 30-40 and so on.  It is tough to predict the age of a person from a single image as perceived age depends upon many factors.

Modules used in this project

OpenCV: As the name suggests, OpenCV is an open-source Computer Vision library. OpenCV is capable of processing real-time images and videos with analytical capabilities. It supports deep learning frameworks like TensorFlow, Pytorch, and Caffe.

Project Implementation

To complete the project, just follow the below steps,

• Detect the faces
• Estimate the age into 8 age groups
• Display the output with an image showing the age group.

SLLATEST
Dataset for age prediction: For this project, we are going to use Adience dataset. It is an open-source dataset made available for the public. This dataset has 26,580 photos and 2,284 numbers of subjects. It already has classified the age groups into 8 ( 0-2, 4-6, 8-13, 15-20, 25-32, 38-43, 48-53, 60-) with gender labels.

• Detect the face: We will detect the face using DNN. We will specifically use the function getFaceBox() to detect the face.
• Now we will load the ageing network and using forward pass to get the age group.
• Show the output of the neural network on the input image and display the image using imshow function.

Observations

It is observed that age groups 0-2, 4-6, 8-13 and 25-32 are predicted with very high accuracy, but the output is heavily biased for age-group 25-32.

The accuracy can be further increased if we use the regression model instead of the classification model, data argumentation and better network architectures. 

Software requirements: Text editor and Python3, Adience dataset.

Programming Languages: Python or C++

SLDYK
Kit required to develop Age Prediction using Image Processing:

• Camera

Technologies you will learn by working on Age Prediction using Image Processing:

• Image Processing
• Computer vision

Any Questions?