Software Projects · Cal Glass
Smart Wearable Glasses · Python, OpenCV, Raspberry Pi · Nov 2021 to Oct 2022
A pair of glasses that watches what you eat and quietly logs it for you. Cal Glass uses a tiny camera and on device computer vision to recognize food in real time, then ships the data to a personal health dashboard so you can finally see the whole loop instead of three disconnected pieces of it.
There was a high school teacher I knew who exercised every single day. He skipped meals constantly. He tried everything. And he was still severely overweight. We watched his effort, we watched the results, and we kept asking why nothing was working. That question became this project.
The honest answer turned out to be structural. The body health loop is broken into pieces that don't talk to each other:
Each piece works in isolation. None of them close the loop. If you want a real plan, you have to shuttle data between three or four tools by hand, and most people give up before they even start.
We dug into the problem and the bottleneck became obvious. Food intake is the hardest data to capture, and it's the one piece every other piece depends on.
Existing food trackers like Lose It and YAZIO are good products, but they're expensive and they ask you to do things humans simply do not do reliably:
The other option is to send a photo to a human fitness advisor and wait for them to interpret it. That works, but it doesn't scale.
There's a Chinese fast food chain called 老乡鸡 (Lao Xiang Ji) where you grab a plastic tray and walk down a line of pottery dishes, picking the things you want. At checkout you slide the tray under a scanner. A camera looks at the plate, identifies every dish, and produces a final price. No barcodes, no manual entry, no human cashier doing the math.
If a restaurant can do plate level recognition automatically at the register, why can't a wearable do the same thing for one person, all day, everywhere they eat?
That was the moment the project clicked. Computer vision was already strong enough. The piece that didn't exist yet was the form factor.
A pair of smart glasses, end to end, hardware and software.
Hardware
Software
The full loop: you eat, the glasses see it, the cloud logs it, the website tells you what to do next. No phone, no manual entry, no remembering to weigh anything.
I led a team of seven as the founder. I designed and 3D printed the frames, assembled the hardware, wrote the detection pipeline, and built the companion website. Across CAD, embedded systems, computer vision, and full stack web, I owned every layer at least once.
This was the first time I ever did 3D printing, the first time I ever wired hardware to software, and the first time I ever shipped a Python project that mattered. It also became my first real taste of research style work: literature review, prototyping, evaluation, writing it up. A lot of what came after, both in my engineering life and my research life, started here.
The natural next step is closing the rest of the loop: integrating real time body data from a wearable (heart rate, activity, sleep) so the recommendations stop being only about food and start being about the whole person.
This is the plain-HTML mirror served to crawlers, LLMs, and curl. Humans with a JavaScript-enabled browser see the rich React/XP-themed SPA at the same URL.