Hue.U

AI-powered skin tone analysis and virtual try-on. Discover your color season from a selfie, then shop and try on outfits matched to your palette — powered by YouCam Skin AI & Apparel VTO.

Comment

Inspiration

We've both stood in a fitting room wondering the same thing: does this color actually suit me? Most people just guess, or they pay for an expensive color consultant to get a real answer. Meanwhile, "seasonal color analysis" — the idea that everyone falls into a Spring, Summer, Autumn, or Winter palette based on their undertone and contrast — has been around since Johannes Itten's color theory and Carole Jackson's Color Me Beautiful in the 1980s. It's real, useful science that most people have never had an accessible way to use. We wanted to put it in everyone's pocket, and pair it with a way to actually act on it: trying clothes on before buying them.

What it does

Hue.U turns a single selfie into a personal color profile and a shopping experience built around it.

  1. Scan — using our own custom in-app camera (not the OS default), with a live face-detection guide so users know exactly when they're positioned correctly before capturing.
  2. Discover your season — YouCam Skin AI analyzes skin, hair, and eye color from the photo and classifies the result into one of four seasons, with a full color palette and a plain-language explanation of why.
  3. Shop matched — our product catalog is filtered and ranked against that palette, so users only see items that are a genuine match.
  4. Try it on — with one more photo, YouCam Apparel VTO generates a realistic virtual try-on, so users can see the fit and color together before ever visiting a fitting room.

How we built it

  • Backend: Node.js/Express, deployed on Heroku, integrating YouCam's V2 Bearer-token API for both Skin AI analysis and Apparel VTO. Color-science logic (undertone classification, contrast calculation, season mapping) is implemented server-side in pure JS.
  • Frontend: React Native / Expo, with a custom-built camera screen (not the OS picker) that runs real-time face detection to guide photo capture, and a glassmorphism-styled UI throughout.
  • Storage: Firebase/Firestore for scan history, Cloudinary for photo hosting.
  • Team split: Alle led the Expo/React Native frontend and UX; Yaya led the Express backend, YouCam API integration, deployment, and product direction.

Challenges we ran into

  • Our first backend deploy crashed in production with a cryptic undefined.replace() error that never worked locally — it turned out PerfectCorp's response doesn't always include every color field, and our code assumed it always would. We fixed the root cause with proper validation and defensive guards instead of patching the symptom.
  • That same assumption broke analysis for hijab-wearing users specifically: PerfectCorp never returns a hair_color when hair isn't visible, which our stricter validation was then wrongly rejecting. We made skin_color the only required field and adapted the contrast calculation and explanation text to work gracefully without hair data — so the app works the same for every user, not just the "default" case.
  • Building a custom camera with real-time face detection (instead of relying on the OS picker) took real trial and error to get right on-device, but it was worth it: users get live feedback instead of repeatedly failing analysis and burning API credits on bad photos.

Accomplishments that we're proud of

  • A fully working end-to-end flow — scan, analysis, matched catalog, and virtual try-on — running live against the real YouCam API, not a mockup.
  • Catching and fixing an inclusivity bug (the hijab/head-covering case) before it ever reached a real user, by treating it as a real requirement rather than an edge case.
  • A custom camera experience with live face-detection guidance, built specifically to reduce failed scans and wasted API calls.

What we learned

That "the API works" and "the API works for everyone" are two different bars, and the second one only gets met if you go looking for the cases your first assumption didn't cover. Also: a lot about defensive backend design — validating what an external API actually returns, not just what its documentation implies it will.

What's next for Hue.U

  • Wiring up src_file_id/ref_file_id chaining so returning users don't need to re-upload photos for every try-on.
  • Expanding the product catalog and refining season-to-product matching.
  • Exploring a 12-season model for more precise recommendations, building on the 4-season foundation we shipped for the hackathon.

Built With

Share this project:

Updates

Submission history