Gender Prediction Using Periocular Region

Project Title

Gender Prediction Using Periocular Region

Project Area of Specialization Artificial IntelligenceProject Summary

Suicidal attacks are becoming common spreading terror regarding veil. There are many cases in recent years where wrong usage of veil result in massive destruction. The expert male suicidal attackers who remain successful from being caught now started using veil; i-e new strategy to divert security concerns pretending themselves as females and targeting our cultural values. For protection of the security system and innocent public who are becoming attacker’s target day by day there is a need for such a system that detect gender even from a certain distance without being focused constantly front of cameras.

Many face detection systems been deployed for security and protection. But there comes a problem when attackers use veil or face covering to hide their identities showing only their “periocular region”. Such reasons compel us to develop a system that detects gender through periocular region. The system we are going to develop will predict gender in a way better than detection systems based on iris. User friendly application where working involves capturing image of user entered, image will be normalized and cropped region of interest then applying machine learning algorithms to identify gender.

If we left with time we will work on “expression reading” from occlude face as well.

Project Objectives

• Efficient recognition system allows user to detect gender
• Handling real world problem by facilitating the user
• Protecting cultural values and promoting usage of veil
• Identifying the people who want such things to happen and safeguard public.
• Using periocular region instead of iris to identify gender from a distance or in crowd.

Project Implementation Method

We will be developing our project in the following modules:

Data Collection:

During this stage, we will train the model using bench marked data set. The model will be evaluated on further data set that we will collect for evaluation of our model.

User Interface:

Developed desktop application will allow the user:

• To capture the image of the person crossed pass-through
• To detect gender

• To record them in database
• Display the gender identified(predicted)

Machine Learning Model:

We will use two strategies in our gender prediction model:

1. We will use CNN with LBP  

And

1. HOG with SVM

And we will compare results from both of these and go with the combination which more accurately predicts gender.

Technologies:

• Xml
• Python
• OpenCV with python

Tools:

• Anaconda distribution
• python GUI framework

Benefits of the Project

• Real world problem regarding security issues will be handled
• Provide security while preserving religious concerns
• Without creating violation we will be able to detect gender through periocular region
• This system will help to secure and promote use of hijab and other cultural values
• Securing educational institutes from suicidal attackers
• Overcome robbery cases where face is partially visible

Technical Details of Final Deliverable

Our system will provide:

• An efficient Desktop application detects gender using periocular region
• Time saving and securing the public from male suicide attackers who used veil for their destructive mission.
• This secure system can be used at security zones

Wireless connected camera with system capture image of person who crossed that security check in. image will send to database or in either case image will be normalized and cropped to extract region of interest. Using CNN with LBP or HOG with SVM results will be attained and compared the one with which will calculate more appropriate will be consider and displayed.

A reliable machine learning model to identify gender from periocular region.

Complete user manual will be available for guidance

Final Deliverable of the Project HW/SW integrated systemType of Industry Security Technologies Artificial Intelligence(AI)Sustainable Development Goals Sustainable Cities and CommunitiesRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	60000
Laptop Core i5 gen or higher	Equipment	1	30000	30000
High quality Camera	Equipment	1	30000	30000