Inspiration

Many people who have unhealthy food cravings can't control themselves and end up overeating, leading to one of two things: Obesity or CHD, both of which are not easily fixable at an older age.

What it does

CalCatcher analyzes the meal image uploaded by the user. It then shows all the foods present along with their respective calories. The user has a history log with all their previous scanned meals as well.

How we built it

I worked on the UI/UX to make it satisfying to click upload, login and register by using smooth fade animations. My teammate worked on the backend (he worked on both technically).

Challenges we ran into

At first the UI looked basic: Simple buttons, no animations, etc. But when I tried adding animations, the animations were either too fast or not showing at all. The loading animation kept showing towards the side of the page regardless of whether or not an image was uploaded. Our biggest challenge was handling the latency and reliability of real-time AI image analysis. Relying solely on external APIs caused delays that ruined the user experience. We also faced significant deployment hurdles with database dependencies (SQLite/Python) on the free-tier cloud infrastructure, which forced us to re-architect our data persistence layer multiple times to ensure stability.

Accomplishments that we're proud of

The construction of a fully functional Frontend and Backend. Working under time pressure. Linking the frontend with the backend and all the testing for both. We are also incredibly proud of the 'Hybrid Intelligence Engine'. The fact that our app can switch between Cloud AI and Local Edge inference instantly ensures zero downtime for the user. We're also proud of the 'Health Score' gamification—we didn't just build a calorie counter; we built a system that interprets data to actually change user behaviour through positive reinforcement.

What we learned

We learned how to implement a 'Circuit Breaker' pattern in backend architecture. Instead of relying 100% on the cloud, we built a system that can degrade gracefully to local heuristics if the API fails. We also learned how to push the limits of Vanilla JavaScript and CSS to create high-performance, 60fps animations (Glassmorphism, Parallax) without the overhead of heavy frontend frameworks like React## What's next for CalCatcher The addition of a subscription feature for users who want to scan more meals per day, or meals for different people. A login system is also in consideration along with a step suggestion to remain healthy.

Share this project:

Updates