Attendance is a critical part of education. The physicality, of course. When I was in college, we had a cut-off of 75%. If the student failed to “attend” 75% of the classes then he/she was debarred from taking the exams. And, we were required to do this for every subject. I wasn’t a good student at all – even by Harry Potter standards – but I was obsessed with answering “yes sir/yes madam” six times a day, six days a week, and I did it for 5 years. My friends were well aware of this frailty in my character and would often gag me when the designated “roll call” took place. It resulted in mayhem, with me – the captive - trying to break free, the teacher gaping at us incredulously, and four burly-looking boys – my friends, the captors - behaving like members of the MOSSAD. Cheap thrill, but immensely entertaining when I see it in hindsight. 

It’s evident, taking attendance manually is cumbersome and time-consuming. To counter these challenges and making every hour spent in school count, the Tamil Nadu e-Governance Agency (TNeGA) has pilot-tested an AI-based facial recognition system and plans are underway to have it implemented in 3000 schools across the state. 

You may well ask, what’s wrong with RFID-based systems that already exist which use physical cards with embedded sensors? Nothing really but these machines require installations that are heavily sensor-based which don’t always give the desired results in the long run and not least, the e-waste one adds by using plastics. Fingerprints-based attendance systems eliminate the need for heavy sensor-based components but there are limitations: only one student can mark attendance at a time, human carelessness (wet/dirty hands, etc.) and, hygiene factors particularly of great significance in COVID times.

This new system employs artificial intelligence, deep learning, and machine learning on top of computer vision, optimized to execute on low-powered Edge Computing devices. There’s something called the Image Analytics Device (IAD), a low-powered Edge Computing device (that I referred to) which captures video feeds from an embedded camera/webcam/IP camera connected over the network. From the captured images, it detects faces, classifies the users, and marks attendance in the database. It can also perform mass identification, thereby saving valuable time. It is mobile-friendly as well and can be integrated with the existing MIS systems. It has brought down the time taken for attendance to one-third of what it was earlier. 

• Data collection phase – a short video of each student with a focus on facial points from all angles, is captured.
• Pre-processing phase – each video file is cropped and cleaned to extract the face thumbnail of each student, and only 50 such images for each student.
• Training phase – the cropped and condensed image data of students are employed to train the deep neural network-based face recognition engine. 
• Evaluation phase – the trained ML model is then evaluated with a set of images the model hasn’t encountered before. The performance of the model is evaluated based on accuracy and recall. By continuous training, an accuracy of 99.6% or thereabouts has been achieved.
• Deployment phase – the model is then deployed on the IAD which recognizes the students and marks attendance. 

The attendance management application is web-based built on Angular framework with SQL database to maintain records. PHP server is used to communicate with the front-end of the database. The system also provides dashboards and the user can navigate to check the day’s/month’s overall attendance or even for a particular student. 

How good is it?

• It can distinguish between identical twins.
• Similar looking siblings.
• Even those with spectacles. 
• It can distinguish students wearing shawls/burkha 
• It can recognize students under varied illumination
• And, even when they aren’t making an eye-contact with the camera. 

The TN government recognizes students who have 100% attendance and gives them a token amount as encouragement. Wherever this system has been adopted there’s been a remarkable improvement in attendance. Habitual latecomers have figured out that there’s no escape or making excuses. 

Technology can be used to develop discipline as we have seen in numerous examples from our daily lives. But this case is a cost-effective way of making the best use of time. And, students are introduced to this facet of technology at a very early age. What stops them from falling in love with technology and marvelling at the realms of possibilities?