Purdue Food Image Dataset
lorenz.ecn.purdue.edu/~vfn — Image dataset for food recognition
← Projects
Project 06
Built
PicNEat
Built · 6 Versions
The Goal
Build my first app — a calorie tracker built specifically for Dining Courts. Point your phone at food, get nutrition data back, and make healthy eating at Purdue feel less like homework.
The Journey — 6 Versions
Started as a terrible little website-looking thing in Summer 2025. Became a cool-looking mockup I'd actually want to use every day. Tried a bunch of AI tools after miserably hating the process of learning Flutter — it worked out well.
V6 Latest
Claude Desktop Version
github.com/mahekparvez/v6picneat
↗
V5
Claude Version
github.com/mahekparvez/v5picneat
↗
V4
Claude Version
github.com/mahekparvez/v4picneat
↗
V3
Image Datasets
github.com/mahekparvez/v3picneat
↗
V2
Lovable Version
github.com/mahekparvez/v2picneat
↗
V1
Flutter Version
github.com/mahekparvez/v1picneat
↗
See It
References
Skillsets Formed
Figma
UI/UX Design
Replit
Lovable
Claude Code
PRD Writing
Hand-Drawing Wireframes
Flutter (basics)
Dart (basics)
Top 3 Learnings
01
Making beautiful UI taught me how to actually build an aesthetic sense — not just copy what looks good, but understand why it works.
02
I don't love the process of coding a frontend, backend, or handling a database. So no fullstack for me — at least not for now.
03
Built my first PRD — wrote it on my iPad. Getting everything out of my head and onto paper before touching code makes everything better.
Mental Note
Personal reminder
Instead of trying to build an app, build something that doesn't involve a backend — backends are messy for non-app-builders. This will likely change in the era of vibe-coding, so keep trying to get better at it with AI.
Grateful For
People who gave feedback, time, or encouragement during the build.
Dr. Fengquing Maggie Zhu
Bruce Coburn
Amanda (Mandy) Kogut
Guatham Vinod
Anna Citrus
Neha Mahesh
Nathaniel Kemme Nash
Rishit Sharma
Jeffrey Tseng
Potential Improvements
Launch on the App Store. $100 was just too much to spend solo — but it should happen eventually.
Got feedback from 3 users. Should have gotten more, listened better, and done it faster. Don't fall in love with your product — fall in love with the impact it has on others.
Make the image detection model sharper. Calorie mapping was working well — the recognition is what needs love.