FTI Students build Facial Recognition System to revamp Attendance

The current pandemic crisis has provided organizations, both large and small, with plenty of opportunities to innovate. As work-from-home becomes a common practice at corporations and academic institutions, alike, it has also become clear that advanced technological solutions are necessary to keep the flow of productivity and efficiency going. 

One such innovative solution has recently been developed by a group of students belonging to the first cohort of the Data Science Certification offered by Frontier Technology Institute (FTI), Pakistan’s premier Data Science Institute in Karachi. The solution is called ‘Real-time Automated Attendance System using Facial Recognition’ which marks attendance of students and employees using face biometrics based on high-definition monitor video and other information technology. 

The need for such a system was identified to reduce the redundant procedures involved in the conventional attendance systems in Pakistan based on a roll call at universities, filling in attendance registers, and fingerprint impressions at organizations. The aim was to develop a contactless, user-friendly, more systematic, and less prone-to-error solution that simplifies the attendance-taking procedure both for the organization and the concerned individual.

FTI students Hassan Ahmed, Hamza Quaid Joher, Rama Abrik, and Hafiz Muhammad Shahid were involved in the project. They were led by Ayman Rehan, in the project management capacity, who is also a faculty member at FTI. The project was divided into two main groups: the back-end and the front-end. The back-end group was responsible for creating the database, data modeling, and programming algorithms. The front-end group designed the user interface, e.g. attendance registry, making the system functional for human use. Hassan and Rama were primarily responsible for handling the back-end processes. Hassan also worked on the front-end graphical user interface (GUI) along with Hamza and Hafiz. However, each student made some contribution, major or minor, to both parts of the project.

The back-end database was created by taking five to ten photos, each, of famous people from Google. The reason for using online photos was to create a large database with many sample pictures to train the system, which improves accuracy. These photos were then fed to a computer that ran a machine-learning algorithm to identify the people and register them as members in the attendance record. As a result of this programming, when a member, whose picture has been recorded, appeared in front of the camera a sixth time, the camera identified her/him automatically – shown by a square appearing in the camera on their face, along with their name as identification. Once identified, this person was marked as ‘present’ on the attendance record. 

To accomplish this project, a two-pronged approach was used in the execution. One was to use a pre-existing Application Programming Interface (API) model and the second was to make the model from scratch. Satisfactory results were obtained through both methods, however, the API method proved to be much more efficient.

“There are many pre-existing API models to choose from,” says Hassan, “most companies prefer to employ these API models since it saves time and they have already been curated for use; only minor tweaks are required to optimize it for your particular project. In our case, the API model gave greater accuracy as compared to the one we made from scratch.”

Rama, who was also involved in the modeling process, described her experience as follows: 

“It is a seemingly daunting task at first to install so many programs and to make sure they are compatible with every team member’s computer. It is also vital that each member has access to the software as eventually, it has to be integrated to perform a single function. However, with the help of my mentors at FTI and the guidance of Ayman, it was made easy. It also proved to be quite fruitful because the API model was perfected by us to get the final result.” 

In order to make an easy-to-use and efficient facial recognition system, a lot of complex work is required at the back-end. Creating the database is just one step; the system must be trained through deep learning in order to function properly. The recognition of faces and simultaneous recording of attendance, as well as the capability to incorporate new members in the attendance system and/or remove a pre-existing member, is a required feature. All corporations or academic institutions which make use of these systems should have the capability to include and exclude people from their attendance records. For this project, after deliberation, it was decided to make use of a Convolutional Neural Network (CNN), hence there were many layers in addition to the input and output layers. The CNN is commonly used to analyze visual imagery and thus it proved to be a wise decision.

Hafiz, who was responsible for creating the picture data set that was used to train the model says, “It was a new skill for me to learn since I had never handled picture data like this before. A hands-on approach allowed me to appreciate the nuances that are present in an apparently simple project.”

A challenge for the whole group was to integrate the back-end with the front-end GUI. A GUI is a user interface that includes graphical elements, such as windows, icons, and buttons. The integration process was quite technical, but an extremely important one, since the back-end effort would be meaningless if an end result wasn’t available for the user to interact with. 

The GUI was made using Python, which is one of the programming languages students are taught at FTI during python certification. The user interface is basically the application available for the end-user. This application contains the entire attendance sheet, which can be accessed to monitor attendance, and edited manually if need be. Other features, such as adding a new member to the attendance sheet, were also made available in the application. 

“The end result, that a user can access, is the application hence the front-end cannot be neglected. The application must be easy-to-use and also efficient in order for people to be interested in using our facial recognition attendance system,” says Hamza. 

In a nutshell, the facial recognition system created is a real-world solution to handle the day-to-day activities of an organization such as a college. The main task is recognizing the faces of the detected person with high accuracy. The system enrolls the subject’s face into the database against the subject’s ID (unique) and name and then marks the attendance of the recognized faces in the database. The automated system maintains the attendance records of students, as manual management of ledgers is a very tedious task. Not only that, manual attendance is prone to human error, which is eliminated in a digital system like this one. 

“It was quite a complicated part of the project; however our project manager and all mentors and teachers at FTI were there to guide us,” describes Hassan, “We have created a very modern technology which can easily be used in schools, colleges, and corporations today.”

Although this facial recognition system is optimized for the attendance of college students it can have numerous other applications. It can also be used in banks, hospitals, supermarkets, factories, or any organization which needs to mark the attendance of its staff. The system developed by the students is not only ingenious in the way how the model, for facial recognition, is optimized for both time and accuracy but also how they are able to come up with an end-to-end solution which included not only the machine learning model but an interactive GUI, a database for storing individual information, and a file which stored the attendance.

The Capstone Project, therefore, gave the students a hands-on experience and equipped them with a real-world understanding of data science – which is going to be an invaluable asset in the near future. Facial recognition systems, like this one, are already in use in many advanced organizations throughout the world. The introduction of such a technology in Pakistan is, therefore, a step towards the digitization of the nation. More importantly, endowing the youth of the country with the skills to develop such a system on their own is nothing short of a breakthrough. 

We hope you found it insightful. To request a demo: info@frontiertechnologyinstitute.com. 

If you would like to discuss ideas, opportunities, and/or corporate training in data science & analytics, please contact us:

· Konain Qurban, CEO, Vancouver/Karachi: konain.qurban@frontiertechnologyinstitute.com

· Behjat Qurban, Managing Director, London: behjat.qurban@frontiertechnologyinstitute.com

· Sana J. Khan, Public Relations Director, Karachi: sana.khan@frontiertechnologyinstitute.com

We aim to bring highly buildable solutions to you and to help you execute your strategies.

Konain Qurban

Konain Qurban is the CEO of FTI. He spent the last 11 years between London and New York, studying and later working in banking, consulting, and non-profit organizations. He earned an MS from Columbia University, New York, and a BS with Honors from City University, London.