When COVID-19 Turned Us Into Accidental Engineers

Covid-19 was terrible, wasn’t it? Back when the pandemic first hit, I was a high school student in Cần Thơ, Vietnam. I still remember it clearly - we were enjoying Lunar New Year break when suddenly a message from our teacher popped up: “Next week we continue the break. There is a disease spreading.” And my first reaction? “Oh wow, extra holiday? Nice.”

Oh how innocent we were.

What we thought would be a short break turned into years of global chaos. Covid-19 affected every aspect of life across the world and completely reshaped how we lived, learned, and interacted. And yes - humanity invented something revolutionary: online learning.

Online classes felt like a gift at first. I would wake up, open my eyes, grab my laptop, and join a Zoom meeting without even changing out of my blanket burrito state. No uniform. No traffic. No early morning rush. Just me, half asleep, pretending to understand calculus. And… well… you can probably guess why my academic performance did not exactly skyrocket during that period. :(

But of course, we could not stay online forever.

Eventually, vaccines were developed. People got vaccinated. The phrase “new normal” entered our daily vocabulary. Schools reopened with new preventive rules. One of them sounded reasonable at first: everyone entering school must have their temperature checked. If you had a fever, you had to go home immediately.

Sounds responsible, right?

Well… here comes the math.

My school had around 2000 students and staff members. We had 2 medical staff and 2 handheld infrared thermometers. Let’s calculate: assume each temperature check takes 5 seconds - and that’s under perfect conditions where every single person lines up nicely like robots and nothing goes wrong. That is 2000 x 5 seconds = 10,000 seconds. That’s nearly 3 hours.

Three hours.

Who has three hours every morning just to enter school? Classes started at 7 AM. It was simply impossible. The policy was good in intention but terrible in execution.

And that was when opportunity knocked.

The Birth of “A Brilliant Project” (According to Two Bored Teenagers)

My best friend and I - I call him Nép (his Facebook name was Phuc Neptune for reasons still unknown to humanity) - were both majoring in Informatics. We needed a project to compete for direct university admission. And most importantly, we were bored out of our minds.

Three perfect conditions.

So we sat in our dormitory room and began thinking about what we dramatically called “a great project.” We were frustrated with the temperature-check situation at school. And suddenly, the idea struck:

Why is temperature measurement not automated?
Why does it waste so much time?

I had read that airports in Vietnam were installing thermal camera systems to monitor incoming passengers. That sparked something in my head. What if we built a similar system for our school? And what if it could also recognize students’ faces to track who entered the campus?

Very efficient.

Very modern.

And definitely not because I wanted a completely legitimate excuse to capture a photo of my crush without suspicion. LOL. Relax, I am not that creepy. I just wanted a nice photo to print and put into her birthday gift box. That’s romantic. Right? Right??

Anyway.

Our Requirements (And One Optional Feature)

We defined our system goals:

Measure body temperature of multiple people simultaneously at the school gate
Detect and identify who that person is
Log all student data into a database
Send alerts to medical staff and to the student if fever is detected
(Optional) Capture a beautiful photo of my crush

Clear. Scientific. Totally professional.

Step One - Thermal Camera Adventures

We searched for thermal cameras and immediately realized they were insanely expensive. So I asked my teacher for help, and we found the MLX90640 - a relatively affordable thermal camera module compatible with Raspberry Pi.

Affordable… is a relative term. We had to skip breakfast for a month to afford it. Raspberry Pi boards were available at school, so we borrowed one.

Back then, AI coding assistants did not exist. We manually read almost every piece of documentation about the MLX90640 just to produce our first thermal image on the Raspberry Pi.

When that first heatmap appeared on the screen, perfectly reading temperature data - we were ecstatic. Step one: success.

Step Two - Face Detection Headaches

For visual recognition, I bought a cheap 720p webcam from Logitech.

Then came the pain.

I did not know Python. I had zero knowledge of computer vision. Machine learning? CNN? Black magic. My teacher connected me with a professor at Cần Thơ University who specialized in artificial intelligence. He talked passionately about convolutional neural networks and deep learning architectures.

I nodded like I understood everything.

I understood nothing.

We were stuck.

A Lucky Break - face-api.js

One day, while desperately searching for solutions, I found face-api.js - a JavaScript API built on top of TensorFlow for browser-based face detection.

I had a strong backend foundation, so I experimented with it. And surprisingly - it worked. I could detect and track one or multiple faces (not identify them yet, but still progress!).

We placed the thermal camera next to the webcam, aligned the frames, and wrote a program so that the webcam would output facial coordinates. Those coordinates were sent to the Raspberry Pi, which then extracted the temperature from the corresponding pixel area on the thermal camera.

And it worked.

It functioned similarly to airport temperature systems.

We were halfway there.

The Recognition Problem

Actual facial recognition - identifying exactly who the person was - remained unsolved. face-api.js was not accurate enough for reliable identification. We struggled with it until the National Science Competition deadline passed.

Time ran out.

So we pivoted.

Instead of perfect identification, we focused on building a monitoring website for school medical staff. The system would:

Display real-time temperature readings
Send alerts when high fever was detected
Capture and send the person’s image to the medical office
Let staff manually identify the student

Was it perfect? No.

Did it work? Surprisingly, yes.

The Result

We competed in ViSEF and won a prize. More importantly, we were granted direct university admission. I remain deeply grateful for that opportunity.

Here is us at Hue - a beautiful city in the Middle of Vietnam

The system was not fully completed. But the lessons were priceless:

Engineering is often about solving 70 percent of a problem effectively
Constraints force creativity
Sometimes, shipping a working imperfect system is better than chasing perfection

Covid-19 is now part of history. But the idea of automated health monitoring systems still feels relevant. The next unexpected global crisis may come when we least expect it.

And maybe next time - we will finish the facial recognition feature.

And just in case you were wondering - no, I do not have a crush on that girl anymore. Somewhere in the middle of debugging thermal pixels, fighting JavaScript face detection, and surviving on instant noodles, I fell for someone else. We started dating during that chaotic project season… and it has been five years since then.

So technically, COVID-19 gave me two things: a science award - and a love story.

Hehe.